US4783803A — Speech recognition apparatus and method

Claim 1

Independent Claim 215 words

A speech recognition system for recognizing a word corresponding to a given utterance spoken after one or more preceding words, said system comprising: means for storing an acoustic model of a given word; means for making a word match score corresponding to the probability that said given word corresponds to acoustic data generated by said utterance, including means for making successive partial comparisons between said acoustic model and said acoustic data and for successively updating said word match score after each such partial comparison; context storing means for storing a language context derived from one or more words spoken prior to said given utterance; language score introducing means, including means for generating a language score for said given word corresponding to a language model probability estimate of the probability that said given word would occur given the language context stored in said context storing means and means for separately altering said word match score after each of a plurality of said partial comparisons by an amount corresponding to said language score; and means for stopping said means for making a word match score from making further partial comparisons between said acoustic model of a given word and said acoustic data when the word match score for that given word is worse than a given threshold.

Claim 2

Dependent Claim 215 words

A speech recognition system for recognizing a word corresponding to a given utterance spoken after one or more preceding words, said system comprising: means for storing an acoustic model of a given word; means for making a word match score corresponding to the probability that said given word corresponds to acoustic data generated by said utterance, including means for making successive partial comparisons between said acoustic model and said acoustic data and for successively updating said word match score after each such partial comparison; context storing means for storing a language context derived from one or more words spoken prior to said given utterance; language score introducing means, including means for generating a language score for said given word corresponding to a language model probability estimate of the probability that said given word would occur given the language context stored in said context storing means and means for separately altering said word match score after each of a plurality of said partial comparisons by an amount corresponding to said language score; and means for stopping said means for making a word match score from making further partial comparisons between said acoustic model of a given word and said acoustic data when the word match score for that given word is worse than a given threshold.

Claim 3

Dependent Claim 75 words

A speech recognition system as described in claim 1, wherein said language score introducing means means for altering said word match score includes means for altering said match score after each of a given number of said partial comparisons and for causing the total amount by which said language score introducing means alters said word match score over all of said given number of partial comparisons to substantially equal the value of said language score.

Claim 4

Dependent Claim 91 words

A speech recognition system as described in claim 1, wherein said word match score is a logarithmic function of an estimate of the probability that said given word corresponds to the acoustic data generated by said utterance and said language score is a logarithmic function of said language model probability estimate, and said means for altering said word match score after each of a plurality of partial comparisons includes means for adding a fraction of said language score to said word match score after each of said plurality of partial comparisons.

Claim 5

Dependent Claim 176 words

A speech recognition method for recognizing a word corresponding to a given utterance spoken after one or more preceding words, said method comprising: storing an acoustic model of a given word; making a word match score corresponding to the probability that said given word corresponds to acoustic data generated by said utterance, including making successive partial comparisons between said acoustic model and said acoustic data and successively updating said word match score after each such partial comparison; storing a language context derived from one or more words spoken prior to said given utterance; generating a language score for said given word corresponding to a language model probability estimate of the probability that said given word would occur given said stored language; separately altering said word match score after each of a plurality of said partial comparisons by an amount corresponding to said language score; and stopping further partial comparisons between said acoustic model of a given word and said acoustic data when the word match score for that given word is worse than a given threshold.

Claim 6

Dependent Claim 176 words

A speech recognition method for recognizing a word corresponding to a given utterance spoken after one or more preceding words, said method comprising: storing an acoustic model of a given word; making a word match score corresponding to the probability that said given word corresponds to acoustic data generated by said utterance, including making successive partial comparisons between said acoustic model and said acoustic data and successively updating said word match score after each such partial comparison; storing a language context derived from one or more words spoken prior to said given utterance; generating a language score for said given word corresponding to a language model probability estimate of the probability that said given word would occur given said stored language; separately altering said word match score after each of a plurality of said partial comparisons by an amount corresponding to said language score; and stopping further partial comparisons between said acoustic model of a given word and said acoustic data when the word match score for that given word is worse than a given threshold.

Claim 7

Dependent Claim 63 words

A speech recognition method as described in claim 4, wherein said altering of said word match score includes altering said match score after each of a given number of said partial comparisons and causing the total amount by which said word match score is altered over all of said given number of partial comparisons to substantially equal the value of said language score.

Claim 8

Dependent Claim 88 words

A speech recognition method as described in claim 4, wherein said word match score is a logarithmic function of an estimate of the probability that said given word corresponds to the acoustic data generated by said utterance and said language score is a logarithmic function of said language model probability estimate, and said altering of said word match score after each of a plurality of partial comparisons includes adding a fraction of said language score to said word match score after each of said plurality of partial comparisons.

Claim 9

Dependent Claim 313 words

A speech recognition system for recognizing a word corresponding to a given utterance spoken after one or more preceding words, said system comprising: means for storing an acoustic model of each of a plurality of words belonging to an originally active vocabulary; currently active vocabulary means for indicating which words in said originally active vocabulary belong to a currently active vocabulary; means for causing said currently active vocabulary means to originally indicate that all the words in said originally active vocabulary are in said currently active vocabulary; means for making a word match score for each word in said originally active vocabulary corresponding to the probability that each such word corresponds to acoustic data generated by said utterance, including means for making a succession of estimation cycles, in each of which, for each word in said currently active vocabulary, a partial comparison is made between the acoustic model of said word and said acoustic data and in each of which said word match score for said word is updated in response to said partial comparison; context storing means for storing a language context derived from one or more words spoken prior to said given utterance; language score introducing means for generating language scores for individual words for which said partial comparisons are made and for altering, during each of a plurality of said estimation cycles, the word match score of each such word by an amount corresponding to the language score of each such word, each of said language scores corresponding to a language model probability estimate of the probability that a given word would occur in the language context stored in said context storing means; and means for causing said currently active vocabulary means to indicate that a given word is no longer in said currently active vocabulary when the word match score for said word is worse than a given threshold.

Claim 10

Dependent Claim 313 words

A speech recognition system for recognizing a word corresponding to a given utterance spoken after one or more preceding words, said system comprising: means for storing an acoustic model of each of a plurality of words belonging to an originally active vocabulary; currently active vocabulary means for indicating which words in said originally active vocabulary belong to a currently active vocabulary; means for causing said currently active vocabulary means to originally indicate that all the words in said originally active vocabulary are in said currently active vocabulary; means for making a word match score for each word in said originally active vocabulary corresponding to the probability that each such word corresponds to acoustic data generated by said utterance, including means for making a succession of estimation cycles, in each of which, for each word in said currently active vocabulary, a partial comparison is made between the acoustic model of said word and said acoustic data and in each of which said word match score for said word is updated in response to said partial comparison; context storing means for storing a language context derived from one or more words spoken prior to said given utterance; language score introducing means for generating language scores for individual words for which said partial comparisons are made and for altering, during each of a plurality of said estimation cycles, the word match score of each such word by an amount corresponding to the language score of each such word, each of said language scores corresponding to a language model probability estimate of the probability that a given word would occur in the language context stored in said context storing means; and means for causing said currently active vocabulary means to indicate that a given word is no longer in said currently active vocabulary when the word match score for said word is worse than a given threshold.

Claim 11

Dependent Claim 76 words

A speech recognition system as described in claim 7, wherein said language score introducing means' means for altering word match scores includes means for altering said match score for a given word during each of a given number of estimation cycles and for causing the amount by which said language score introducing means alters said word match score over all of said given number of estimation cycles to substantially equal the value of said language score.

Claim 12

Dependent Claim 103 words

A speech recognition system as described in claim 7, wherein said word match score for a given word is a logarithmic function of an estimate of the probability that said given word corresponds to the acoustic data generated by said utterance and said language score for said given word is a logarithmic function of said language model probability estimate for said given word, and said means for altering said word match score after each of a plurality of partial comparisons includes means for adding a fraction of said language score to said word match score after each of said plurality of partial comparisons.

Claim 13

Dependent Claim 86 words

A speech recognition system as described in claim 9, wherein said language score introducing means' means for altering word match scores includes means for altering said word match score for a given word for each of a given number of estimation cycles and for causing the amount by which said language score introducing means alters said word match score during each of said given number of estimation cycles to correspond to the value of said language score for said given word divided by said given number.

Claim 14

Dependent Claim 83 words

A speech recognition system as described in claim 7, wherein said language score introducing means' means for altering said word match score for a given word during a given estimation cycle includes means for altering said match score by an amount which is normalized for each given language context stored in said context storing means so that the language scores corresponding to the lowest likelihood language model probability estimates in each language context have approximately the same effect on said word match scores.

Claim 15

Dependent Claim 73 words

A speech recognition system as described in claim 7, further including means for receiving said acoustic data generated by said utterance as a sequence of frames, each of which includes a vector of parameters indicating acoustic properties of the utterance during a given period of time; and wherein said means for making a succession of estimation cycles includes means for making a separate estimation cycle for each of a succession of said frames.

Claim 16

Dependent Claim 312 words

A speech recognition system as described in claim 12, wherein: said means for storing an acoustic model of each of a plurality of words includes means for storing for each such acoustic model a sequence of nodes, each of which has a vector of parameters corresponding to that of said frames; said means for making a word match score includes dynamic programming means for seeking to optimally match a succession of said frames against the sequence of nodes in the acoustic model of each word in said originally active vocabulary, and said means for making a succession of estimation cycles includes means for making, for each of a plurality of nodes in the acoustic model of each word in said currently active vocabulary, during each estimation cycle, a comparison between said node and a given frame, means for using said comparison to update a node match score associated with each such node, each such node match score being a function of the probability of an optimal dynamic programming path to said node through the previous nodes of its corresponding acoustic model, and means for selecting the most probable node match score for a given word as its word match score, and said language score introducing means including means for storing, for each of said plurality of nodes associated with each word in said currently active vocabulary, a word start value which indicates at what frame, according to the optimal path associated with said node, the utterance of said word began, means for altering, during each of a plurality of estimation cycles, the node match score of each of said plurality of nodes by an amount corresponding to the language score of each such node's associated word, and means for using said word start value to determine during which plurality of estimation cycles to alter the node match score associated with said node.

Claim 17

Dependent Claim 84 words

A speech recognition system as described in claim 8, wherein said means for causing said currently active vocabulary means to indicate that a given word is no longer in said currently active vocabulary when the word match score for said word is worse than a given threshold includes means for varying the value of said threshold relative to said word match score as a function of changes in the word match score generated for one or more other words in said currently active vocabulary.

Claim 18

Dependent Claim 56 words

A speech recognition system as described in claim 14, wherein said means for varying said threshold includes means for setting the value of said threshold relative to said word match scores as a function of the value of the word match score indicating the greatest likelihood of occurrence of any word in the currently active vocabulary.

Claim 19

Dependent Claim 268 words

A speech recognition method for recognizing a word corresponding to a given utterance spoken after one or more preceding words, said method comprising: storing an acoustic model of each of a plurality of words belonging to an originally active vocabulary; indicating which words in said originally active vocabulary belong to a currently active vocabulary; originally indicating that all the words in said originally active vocabulary are in said currently active vocabulary; making a word match score for each word in said originally active vocabulary corresponding to the probability that each such word corresponds to acoustic data generated by said utterance, including making a succession of estimation cycles, in each of which, for each word in said currently active vocabulary, a partial comparison is made between the acoustic model of said word and said acoustic data and in each of which said word match score for said word is updated in response to said partial comparison; storing a language context derived from one or more words spoken prior to said given utterance; generating language scores for individual words for which said partial comparisons are made; altering, during each of a plurality of said estimation cycles, the word match score of each such word by an amount corresponding to the language score of each such word, each of said language scores corresponding to a language model probability estimate of the probability that a given word would occur in said stored language context; and indicating that a given word is no longer in said currently active vocabulary when the word match score for said word is worse than a given threshold.

Claim 20

Dependent Claim 268 words

A speech recognition method for recognizing a word corresponding to a given utterance spoken after one or more preceding words, said method comprising: storing an acoustic model of each of a plurality of words belonging to an originally active vocabulary; indicating which words in said originally active vocabulary belong to a currently active vocabulary; originally indicating that all the words in said originally active vocabulary are in said currently active vocabulary; making a word match score for each word in said originally active vocabulary corresponding to the probability that each such word corresponds to acoustic data generated by said utterance, including making a succession of estimation cycles, in each of which, for each word in said currently active vocabulary, a partial comparison is made between the acoustic model of said word and said acoustic data and in each of which said word match score for said word is updated in response to said partial comparison; storing a language context derived from one or more words spoken prior to said given utterance; generating language scores for individual words for which said partial comparisons are made; altering, during each of a plurality of said estimation cycles, the word match score of each such word by an amount corresponding to the language score of each such word, each of said language scores corresponding to a language model probability estimate of the probability that a given word would occur in said stored language context; and indicating that a given word is no longer in said currently active vocabulary when the word match score for said word is worse than a given threshold.

Claim 21

Dependent Claim 67 words

A speech recognition method as described in claim 16, wherein said altering of word match scores includes altering said match score for a given word during each of a given number of estimation cycles and for causing the amount by which said language score alters said word match score over all of said given number of estimation cycles to substantially equal the value of said language score.

Claim 22

Dependent Claim 100 words

A speech recognition method as described in claim 16, wherein said word match score for a given word is a logarithmic function of an estimate of the probability that said given word corresponds to the acoustic data generated by said utterance and said language score for said given word is a logarithmic function of said language model probability estimate for said given word, and said altering of said word match score after each of a plurality of partial comparisons includes adding a fraction of said language score to said word match score after each of said plurality of partial comparisons.

Claim 23

Dependent Claim 77 words

A speech recognition method as described in claim 18, wherein said altering of word match scores includes altering said word match score for a given word for each of a given number of estimation cycles and for causing the amount by which said language score alters said word match score during each of said given number of estimation cycles to correspond to the value of said language score for said given word divided by said given number.

Claim 24

Dependent Claim 71 words

A speech recognition method as described in claim 16, wherein said altering of said word match score for a given word during a given estimation cycle include altering said match score by an amount which is normalized for each given stored language context so that the language scores corresponding to the lowest likelihood language model probability estimates in each language context have approximately the same effect on said word match scores.

Claim 25

Dependent Claim 68 words

A speech recognition method as described in claim 16: further including receiving said acoustic data generated by said utterance as a sequence of frames, each of which includes a vector of parameters indicating acoustic properties of the utterance during a given period of time; and wherein said making of a succession of estimation cycles includes making a separate estimation cycle for each of a succession of said frames.

Claim 26

Dependent Claim 301 words

A speech recognition method as described in claim 21, wherein: said storing of an acoustic model of each of a plurality of words includes storing for each such acoustic model a sequence of nodes, each of which has a vector of parameters corresponding to that of said frames; said making of a word match score includes seeking through the use of dynamic programming to optimally match a succession of said frames against the sequence of nodes in the acoustic model of each word in said originally active vocabulary, and said making of a succession of estimation cycles includes making, for each of a plurality of nodes in the acoustic model of each word in said currently active vocabulary, during each estimation cycle, a comparison between said node and a given frame, using said comparison to update a node match score associated with each such node, each such node match score being a function of the probability of an optimal dynamic programming path to said node through the previous nodes of its corresponding acoustic model, and selecting the most probable node match score for a given word as its word match score, and said altering of the word match score of each word includes storing, for each of said plurality of nodes associated with each word in said currently active vocabulary, a word start value which indicates at what frame, according to said optimal path to said node, the utterance of said word began; altering, during each of a plurality of estimation cycles, the node match score of each of said plurality of nodes by an amount corresponding to the language score of each such node's associated word; and using said word start value to determine during which plurality of estimation cycles to alter the node match score associated with said node.

Claim 27

Dependent Claim 73 words

A speech recognition method as described in claim 16, wherein said indicating that a given word is no longer in said current active vocabulary when the word match score for said word is worse than a given threshold includes varying the value of said threshold relative to said word match score as a function of changes in the word match score generated for one or more other words in said currently active vocabulary.

Claim 28

Dependent Claim 53 words

A speech recognition method as described in claim 23, wherein said varying of said threshold includes setting the value of said threshold relative to said word match scores as a function of the value of the word match score indicating the greatest likelihood of occurrence of any word in the currently active vocabulary.

Claim 29

Dependent Claim 209 words

A pattern recognition system for recognizing an individual pattern in a collection of data given a context of one or more other patterns which have been previously identified, said system comprising: means for storing a model of a given pattern; means for making a pattern match score corresponding to the probability that said given pattern corresponds to said collection of data, including means for making successive partial comparisons between said pattern's model and said data and for successively updating said pattern match score after each such partial comparison; context storing means for storing a context derived from one or more patterns which have been previously identified; context score introducing means, including means for generating a context score for said given pattern corresponding to a context probability estimate of the probability that said given pattern would occur given the context stored in said context storing means and means for separately altering said pattern match score after each of a plurality of said partial comparison by an amount corresponding to said context score; and means for stopping said means for making a pattern match score from making further partial comparisons between said model and said data when the pattern match score for that given pattern is worse than a given threshold.

Claim 30

Dependent Claim 209 words

A pattern recognition system for recognizing an individual pattern in a collection of data given a context of one or more other patterns which have been previously identified, said system comprising: means for storing a model of a given pattern; means for making a pattern match score corresponding to the probability that said given pattern corresponds to said collection of data, including means for making successive partial comparisons between said pattern's model and said data and for successively updating said pattern match score after each such partial comparison; context storing means for storing a context derived from one or more patterns which have been previously identified; context score introducing means, including means for generating a context score for said given pattern corresponding to a context probability estimate of the probability that said given pattern would occur given the context stored in said context storing means and means for separately altering said pattern match score after each of a plurality of said partial comparison by an amount corresponding to said context score; and means for stopping said means for making a pattern match score from making further partial comparisons between said model and said data when the pattern match score for that given pattern is worse than a given threshold.

Claim 31

Dependent Claim 74 words

A pattern recognition system as described in claim 25, wherein said context score introducing means' means for altering said pattern match score includes means for altering said match score after each of a given number of partial comparisons and for causing the total amount by which said context score introducing means alters said pattern match score over all of said given number of partial comparisons to substantially equal the value of said context score.

Claim 32

Dependent Claim 87 words

A pattern recognition system as described in claim 25, wherein said pattern match score is a logarithmic function of an estimate of the probability that said given pattern corresponds to said collection of data and said context score is a logarithmic function of said context probability estimate, and said means for altering said pattern match score after each of a plurality of partial comparisons includes means for adding a fraction of said context score to said pattern match score after each of said plurality of partial comparisons.

Claim 33

Dependent Claim 172 words

A pattern recognition method for recognizing an individual pattern in a collection of data given a context of one or more other patterns which have been previously identified, said method comprising: storing a model of a given pattern; making a pattern match score corresponding to the probability that said given pattern corresponds to said collection of data, including making successive partial comparisons between said pattern's model and said data and for successively updating said pattern match score after each such partial comparison; storing a context derived from one or more patterns which have been previously identified; generating a context score for said given pattern corresponding to a context probability estimate of the probability that said given pattern would occur given said stored context; separately altering said pattern match score after each of a plurality of said partial comparison by an amount corresponding to said context score; and stopping further partial comparisons between said model and said data when the pattern match score for that given pattern is worse than a given threshold.

Claim 34

Dependent Claim 172 words

A pattern recognition method for recognizing an individual pattern in a collection of data given a context of one or more other patterns which have been previously identified, said method comprising: storing a model of a given pattern; making a pattern match score corresponding to the probability that said given pattern corresponds to said collection of data, including making successive partial comparisons between said pattern's model and said data and for successively updating said pattern match score after each such partial comparison; storing a context derived from one or more patterns which have been previously identified; generating a context score for said given pattern corresponding to a context probability estimate of the probability that said given pattern would occur given said stored context; separately altering said pattern match score after each of a plurality of said partial comparison by an amount corresponding to said context score; and stopping further partial comparisons between said model and said data when the pattern match score for that given pattern is worse than a given threshold.

Claim 35

Dependent Claim 62 words

A pattern recognition method as described in claim 28, wherein said altering of said pattern match score includes altering said match score after each of a given number of partial comparisons and causing the total amount by which said pattern match score is altered over all of said given number of partial comparisons to substantially equal the value of said context score.

Claim 36

Dependent Claim 84 words

A pattern recognition method as described in claim 28, wherein said pattern match score is a logarithmic function of an estimate of the probability that said given pattern corresponds to said collection of data and said context score is a logarithmic function of said context probability estimate, and said altering of said pattern match score after each of a plurality of partial comparisons includes adding a fraction of said context score to said pattern match score after each of said plurality of partial comparisons.

Claim 37

Dependent Claim 311 words

A pattern recognition system for recognizing an individual pattern in a collection of data given a context of one or more other patterns which have been previously identified, said system comprising: means for storing a model of each of a plurality of patterns belonging to an originally active set of patterns; currently active set means for indicating which patterns in said originally active set belong to a currently active set of patterns; means for causing said currently active set means to originally indicate that all the patterns in said originally active set are in said currently active set; means for making a pattern match score corresponding to the probability that each pattern in said originally active set occurs in said collection of data, including means for making a succession of estimation cycles, in each of which, for each pattern in said currently active set, a partial comparison is made between the model of said pattern and said collection of data and in each of which said pattern match score for said pattern is updated in response to said partial comparison; context storing means for storing a context derived from one or more patterns previously identified; context score introducing means for generating context scores for patterns for which said partial comparisons are made and for altering, during each of a plurality of said estimation cycles, the pattern match score of each such pattern by an amount which corresponds to the context score of each such pattern, each of said context scores corresponding to a context probability estimate of the probability that its associated pattern would occur in the context stored in said context storing means; and means for causing said currently active set means to indicate that a given pattern is no longer in said currently active set when the pattern match score for said pattern is worse than a given threshold.

Claim 38

Dependent Claim 311 words

A pattern recognition system for recognizing an individual pattern in a collection of data given a context of one or more other patterns which have been previously identified, said system comprising: means for storing a model of each of a plurality of patterns belonging to an originally active set of patterns; currently active set means for indicating which patterns in said originally active set belong to a currently active set of patterns; means for causing said currently active set means to originally indicate that all the patterns in said originally active set are in said currently active set; means for making a pattern match score corresponding to the probability that each pattern in said originally active set occurs in said collection of data, including means for making a succession of estimation cycles, in each of which, for each pattern in said currently active set, a partial comparison is made between the model of said pattern and said collection of data and in each of which said pattern match score for said pattern is updated in response to said partial comparison; context storing means for storing a context derived from one or more patterns previously identified; context score introducing means for generating context scores for patterns for which said partial comparisons are made and for altering, during each of a plurality of said estimation cycles, the pattern match score of each such pattern by an amount which corresponds to the context score of each such pattern, each of said context scores corresponding to a context probability estimate of the probability that its associated pattern would occur in the context stored in said context storing means; and means for causing said currently active set means to indicate that a given pattern is no longer in said currently active set when the pattern match score for said pattern is worse than a given threshold.

Claim 39

Dependent Claim 76 words

A pattern recognition system as described in claim 31, wherein said context score introducing means' means for altering pattern match scores includes means for altering said match score for a given pattern during each of a given number of estimation cycles and for causing the amount by which said context score introducing means alters said pattern match score over all of said given number of estimation cycles to substantially equal the value of said language score.

Claim 40

Dependent Claim 98 words

A pattern recognition system as described in claim 31, wherein said match score for a given pattern is a logarithmic function of an estimate of the probability that said given pattern occurs in said collection of data and said context score for said given pattern is a logarithmic function of said context probability estimate for said given pattern, and said means for altering said pattern match score after each of a plurality of partial comparisons includes means for adding a fraction of said context score to said pattern match score after each of said plurality of partial comparisons.

Claim 41

Dependent Claim 85 words

A pattern recognition system as described in claim 33, wherein said context score introducing means' means for altering pattern match scores includes means for altering said match score for a given pattern for each of a given number of estimation cycles and for causing the amount by which said context score introducing means alters said pattern match score during each of said given number of estimation cycles to correspond to the value of said context score for said given word divided by said given number.

Claim 42

Dependent Claim 83 words

A pattern recognition system as described in claim 31, wherein said means for causing said currently active set means to indicate that a given pattern is no longer in said currently active set when the pattern match score for said pattern becomes worse than a given threshold includes means for varying the value of said threshold relative to said pattern match scores as a function of changes in the match score generated for one or more other patterns in said currently active set.

Claim 43

Dependent Claim 56 words

A pattern recognition system as described in claim 35, wherein said means for varying said threshold includes means for setting the value of said threshold relative to said pattern match score as a function of the value of the pattern match score indicating the greatest likelihood of occurrence of any pattern in the currently active set.

Claim 44

Dependent Claim 273 words

A pattern recognition method for recognizing an individual pattern in a collection of data given a context of one or more other patterns which have been previously identified, said method comprising: storing a model of each of a plurality of patterns belonging to an originally active set of patterns; indicating which patterns in said originally active set belong to a currently active set of patterns; causing said currently active set means to originally indicate that all the patterns in said originally active set are in said currently active set; making a pattern match score corresponding to the probability that each pattern in said originally active set occurs in said collection of data, including making a succession of estimation cycles, in each of which, for each pattern in said currently active set, a partial comparison is made between the model of said pattern and said collection of data and in each of which said pattern match score for said pattern is updated in response to said partial comparison; storing a context derived from one or more patterns previously identified; generating context scores for patterns for which said partial comparisons are made; altering, during each of a plurality of said estimation cycles, the pattern match score of each such pattern by an amount which corresponds to the context score of each such pattern, each of said context scores corresponding to a context probability estimate of the probability that its associated pattern would occur in said stored context; and indicating that a given pattern is no longer in said currently active set when the pattern match score for said pattern is worse than a given threshold.

Claim 45

Dependent Claim 273 words

A pattern recognition method for recognizing an individual pattern in a collection of data given a context of one or more other patterns which have been previously identified, said method comprising: storing a model of each of a plurality of patterns belonging to an originally active set of patterns; indicating which patterns in said originally active set belong to a currently active set of patterns; causing said currently active set means to originally indicate that all the patterns in said originally active set are in said currently active set; making a pattern match score corresponding to the probability that each pattern in said originally active set occurs in said collection of data, including making a succession of estimation cycles, in each of which, for each pattern in said currently active set, a partial comparison is made between the model of said pattern and said collection of data and in each of which said pattern match score for said pattern is updated in response to said partial comparison; storing a context derived from one or more patterns previously identified; generating context scores for patterns for which said partial comparisons are made; altering, during each of a plurality of said estimation cycles, the pattern match score of each such pattern by an amount which corresponds to the context score of each such pattern, each of said context scores corresponding to a context probability estimate of the probability that its associated pattern would occur in said stored context; and indicating that a given pattern is no longer in said currently active set when the pattern match score for said pattern is worse than a given threshold.

Claim 46

Dependent Claim 67 words

A pattern recognition method as described in claim 37, wherein said altering of pattern match scores includes altering said match score for a given pattern during each of a given number of estimation cycles and for causing the amount by which said context score alters said pattern match score over all of said given number of estimation cycles to substantially equal the value of said language score.

Claim 47

Dependent Claim 94 words

A pattern recognition method as described in claim 37, wherein said match score for a given pattern is a logarithmic function of an estimate of the probability that said given pattern occurs in said collection of dta and said context score for said given pattern is a logarithmic function of said context probability estimate for said given pattern, and said altering said pattern match score after each of a plurality of partial comparisons includes adding a fraction of said context score to said pattern match score after each of said plurality of partial comparisons.

Claim 48

Dependent Claim 76 words

A pattern recognition method as described in claim 39, wherein said altering of pattern match scores includes altering said match score for a given pattern for each of a given number of estimation cycles and for causing the amount by which said context score alters said pattern match score during each of said given number of estimation cycles to correspond to the value of said context score for said given word divided by said given number.

Claim 49

Dependent Claim 72 words

A pattern recognition method ass described in claim 37, wherein said indicating that a given pattern is no longer in said currently active set when the pattern match score for said pattern becomes worse than a given threshold includes varying the value of said threshold relative to said pattern match scores as a function of changes in the match score generated for one or more other patterns in said currently active set.

Claim 50

Dependent Claim 53 words

A pattern recognition method as described in claim 41, wherein said varying of said threshold includes setting the value of said threshold relative to said pattern match score as a function of the value of the pattern match score indicating the greatest likelihood of occurrence of any pattern in the currently active set.

Claim 51

Dependent Claim 120 words

A speech recognition system for recognizing a word corresponding to a given spoken utterance, said system comprising: means for making, for each word of an initial vocabulary, a relatively rapid scoring, which produces a rapid match score corresponding to the probability that said word corresponds to a sequence of acoustic data generated by said utterance; means for placing a word from said initial vocabulary into an originally active vocabulary when said rapid match score for said word is better than a given threshold; and means for making, for each word of said originally active vocabulary, a more lengthy scoring, which produces a more accurate match score corresponding to the probability that such word corresponds to said sequence of acoustic data.

Claim 52

Dependent Claim 120 words

A speech recognition system for recognizing a word corresponding to a given spoken utterance, said system comprising: means for making, for each word of an initial vocabulary, a relatively rapid scoring, which produces a rapid match score corresponding to the probability that said word corresponds to a sequence of acoustic data generated by said utterance; means for placing a word from said initial vocabulary into an originally active vocabulary when said rapid match score for said word is better than a given threshold; and means for making, for each word of said originally active vocabulary, a more lengthy scoring, which produces a more accurate match score corresponding to the probability that such word corresponds to said sequence of acoustic data.

Claim 53

Dependent Claim 54 words

A speech recognition system as described in claim 43, wherein both said means for making a rapid scoring and said means for making a lengthy scoring include means for calculating an acoustic likelihood score indicating the closeness of a match between an acoustic model of a given word and said sequence of acoustic data.

Claim 54

Dependent Claim 76 words

A speech recognition system as described in claim 44, wherein both said means for making a rapid scoring and said means for making a lengthy scoring include means for calculating their scores for a given word based in part on a language score which corresponds to a language model probability estimate of the probability that said given word would occur given a language context derived from one or more words which have preceded said given word.

Claim 55

Dependent Claim 44 words

A speech recognition system as described in claim 44, fuerther including means for storing said sequence of acoustic data as a sequence of frames, each of which includes a vector of parameters indicating acoustic properties of the utterance during a given period of time.

Claim 56

Dependent Claim 56 words

A speech recognition system as described in claim 46, wherein said means for making a rapid scoring includes means for making said scoring based on a number of said recorded frames which is less than the number of said frames which occur normally during the utterance of a majority of the words in said initial vocabulary.

Claim 57

Dependent Claim 58 words

A speech recognition system as described in claim 46, wherein said means for making a rapid scoring includes means for averaging corresponding acoustic parameters in a plurality of said frames and means for comparing the closeness of the resulting set of averaged parameters with a corresponding set of averaged parameters associated with each word in said initial vocabulary.

Claim 58

Dependent Claim 74 words

A speech recognition system as described in claim 48, wherein said means for making a rapid scoring includes means for causing said averaging means to separately perform its averaging for each of a plurality of sequences of frames and for causing said means for comparing to compare the closeness of each of the resulting sets of averaged parameters with each of a corresponding set of parameters associated with each word in said initial vocabulary.

Claim 59

Dependent Claim 50 words

A speech recognition systems as described in claim 49, wherein said means for causing said averaging means to perform its averaging separately for each of a plurality of sequences of frames including means for causing at least two of said sequences to include one or more of the same frames.

Claim 60

Dependent Claim 229 words

A speech recognition system as described in claim 43, wherein said means for making a rapid scoring includes means for making said scoring for each word in said initial vocabulary, one word at a time and wherein said means for making a lengthy scoring includes the following: means for storing an acoustic model of each of a plurality of words belonging to said originally active vocabulary; currently active vocabulary means for indicating which words in said originally active vocabulary belong to a currently active vocabulary; means for causing said currently active vocabulary means to originally indicate that all the words in said originally active vocabulary are in said currently active vocabulary; means for making a word match score corresponding to the probability that each word in said originally active vocabulary corresponds to acoustic data generated by said utterance, including means for making a succession of estimation cycles, in each of which, for each word in said currently active vocabulary, a partial comparison is made between the acoustic model of said word and said acoustic data and in which said match score for said word is updated in response to said partial comparison; and means for causing said currently active vocabulary means to indicate that a given word is no longer in said currently active vocabulary when the word match score for said word is worse than a given threshold.

Claim 61

Dependent Claim 92 words

A speech recognition system as described in claim 51, further including means for storing said sequence of acoustic data as a sequence of frames, each of which includes a vector of parameters indicating acoustic properties of the utterance during a given period of time; and wherein said means for making a succession of estimation cycles includes means for making one such cycle for each of a succession of said frames and for making a comparison of the acoustic model of each of a plurality of words with said frame during said cycle.

Claim 62

Dependent Claim 84 words

A speech recognition system as described in claim 52, further including: means for producing a start score corresponding to an estimated probability that the utterance of a word has not begun at a given point in said sequence of recorded frames; and means for resetting said currently active vocabulary to include all the words contained in said originally active vocabulary when, after one or more estimation cycles, said means for producing a start score produces a start score which is better than a threshold.

Claim 63

Dependent Claim 106 words

A speech recognition system as described in claim 53, wherein: said currently active vocabulary means includes means for keeping all the words in the currently active vocabulary as a linked list; said means for causing said currently active vocabulary means to indicate that a given word is no longer in said currently active vocabulary includes means for causing said word which is no longer to be in said currently active vocabulary to be un-linked from said linked list; and said means for resetting said currently active vocabulary includes means for causing all the words in said originally active vocabulary to be re-linked into said linked list.

Claim 64

Dependent Claim 131 words

A speech recognition system as described in claim 54, further including means for representing said initial vocabulary as a list of word records, each of which has a means for storing a pointer to another word record in said list; and said means for causing said currently active vocabulary means to originally indicate that all the words in said originally active vocabulary are in said currently active vocabulary includes means for pointing to a word record which represents a first word in said originally active vocabulary and for causing the means for storing a pointer in each word record associated with a word in said originally active vocabulary to point to the word record associated with the next word in said originally active vocabulary when there is such a next word.

Claim 65

Dependent Claim 114 words

A speech recognition method for recognizing a word corresponding to a given spoken utterance, said method comprising: making, for each word of an initial vocabulary, a relatively rapid scoring, which produces a rapid match score corresponding to the probability that said word corresponds to a sequence of acoustic data generated by said utterance; placing a word from said initial vocabulary into an originally active vocabulary when said rapid match score for said word is better than a given threshold; and making, for each word of said originally active vocabulary, a more lengthy scoring, which produces a more accurate match score corresponding to the probability that such word corresponds to said sequence of acoustic data.

Claim 66

Dependent Claim 114 words

A speech recognition method for recognizing a word corresponding to a given spoken utterance, said method comprising: making, for each word of an initial vocabulary, a relatively rapid scoring, which produces a rapid match score corresponding to the probability that said word corresponds to a sequence of acoustic data generated by said utterance; placing a word from said initial vocabulary into an originally active vocabulary when said rapid match score for said word is better than a given threshold; and making, for each word of said originally active vocabulary, a more lengthy scoring, which produces a more accurate match score corresponding to the probability that such word corresponds to said sequence of acoustic data.

Claim 67

Dependent Claim 50 words

A speech recognition method as described in claim 56, wherein both said making of a rapid scoring and said making of a lengthy scoring include calculating an acoustic likelihood score indicating the closeness of a match between an acoustic model of a given word and said sequence of acoustic data.

Claim 68

Dependent Claim 72 words

A speech recognition method as described in claim 57, wherein both said making of a rapid scoring and said making of a lengthy scoring include calculating their scores for a given word based in part on a language score which corresponds to a language model probability estimate of the probability that said given word would occur given a language context derived from one or more words which have preceded said given word.

Claim 69

Dependent Claim 42 words

A speech recognition method as described in claim 57, further including storing said sequence of acoustic data as a sequence of frames, each of which includes a vector of parameters indicating acoustic properties of the utterance during a given period of time.

Claim 70

Dependent Claim 53 words

A speech recognition method as described in claim 59, wherein said making of a rapid scoring includes making said scoring based on a number of said recorded frames which is less than the number of said frames which occur normally during the utterance of a majority of the words in said initial vocabulary.

Claim 71

Dependent Claim 53 words

A speech recognition method as described in claim 59, wherein said making of a rapid scoring includes averaging corresponding acoustic parameters in a plurality of said frames and comparing the closeness of the resulting set of averaged parameters with a corresponding set of averaged parameters associated with each word in said initial vocabulary.

Claim 72

Dependent Claim 69 words

A speech recognition method as described in claim 61, wherein said making of a rapid scoring includes causing said averaging means to separately perform its averaging for each of a plurality of sequences of frames and for causing said comparing to compare the closeness of each of the resulting sets of averaged parameters with each of a corresponding set of parameters associated with each word in said initial vocabulary.

Claim 73

Dependent Claim 47 words

A speech recognition method as described in claim 62, wherein said causing of said averaging means to perform its averaging separately for each of a plurality of sequences of frames including causing at least two of said sequences to include one or more of the same frames.

Claim 74

Dependent Claim 196 words

A speech recognition method as described in claim 56, wherein said making of a rapid scoring includes making said scoring for each word in said initial vocabulary, one word at a time and wherein said making of a lengthy scoring includes the following: storing an acoustic model of each of a plurality of words belonging to said originally active vocabulary; indicating which words in said originally active vocabulary belong to a currently active vocabulary; originally indicating that all the words in said originally active vocabulary are in said currently active vocabulary; making a word match score corresponding to the probability that each word in said originally active vocabulary corresponds to acoustic data generated by said utterance, including making a succession of estimation cycles, in each of which, for each word in said currently active vocabulary, a partial comparison is made between the acoustic model of said word and said acoustic data and in which said match score for said word is updated in response to said partial comparison; and indicating that a given word is no longer in said currently active vocabulary when the word match score for said word is worse than a given threshold.

Claim 75

Dependent Claim 87 words

A speech recognition method as described in claim 64, further including storing said sequence of acoustic data as a sequence of frames, each of which includes a vector of parameters indicating acoustic properties of the utterance during a given period of time; and wherein said making of a succession of estimation cycles includes making one such cycle for each of a succession of said frames and for making a comparson of the acoustic model of each of a plurality of words with said frame during said cycle.

Claim 76

Dependent Claim 79 words

A speech recognition method as described in claim 65, further including: producing a start score corresponding to an estimated probability that the utterance of a word has not begun at a given point in said sequence of recorded frames; and resetting said currently active vocabulary to include all the words contained in said originally active vocabulary when, after one or more estimation cycles, said producing of a start score produces a start score which is better than a threshold.

Claim 77

Dependent Claim 101 words

A speech recognition method as described in claim 66, wherein: said indicating of which words in said originally active vocabulary belong to a currently active vocabulary includes keeping all the words in the currently active vocabulary as a linked list; said indicating that a given word is no longer in said currently active vocabulary includes causing said word which is no longer to be in said currently active vocabulary to be unlinked from said linked list; and said resetting of said currently active vocabulary includes causing all the words in said originally active vocabulary to be re-linked into said linked list.

Claim 78

Dependent Claim 109 words

A speech recognition method as described in claim 67, further including representing said initial vocabulary as a list of word records, each of which has a pointer to another word record in said list; and said originally indicating that all the words in said originally active vocabulary are in said currently active vocabulary includes pointing to a word record which represents a first word in said originally active vocabulary and causing said pointer in each word record associated with a word in said originally active vocabulary to point to the word record associated with the next word in said originally active vocabulary when there is such a next word.

Claim 79

Dependent Claim 159 words

A speech recognition system for recognizing a word in a sequence of acoustic data, said system comprising: means for storing an acoustic model of expected background noise; means for comparing said acoustic model with successive portions of said sequence of acoustic data and producing a silence score corresponding to the probability that such successive portions of data correspond to background noise and thus that said utterance has not yet begun; means for determining from said acoustic data the acoustic amplitude corresponding to successive portions of said data; means for producing an accumulated sound score which corresponds a sum of to the amount of said acoustic amplitude over said successive portions of data, said accumulated sound score inversely corresponding to the probability that an utterance has not yet begun; and means, responsive to both said silence score and said accumulated sound score for determining during which of said successive portions of data to look for the start of said utterance.

Claim 80

Dependent Claim 159 words

A speech recognition system for recognizing a word in a sequence of acoustic data, said system comprising: means for storing an acoustic model of expected background noise; means for comparing said acoustic model with successive portions of said sequence of acoustic data and producing a silence score corresponding to the probability that such successive portions of data correspond to background noise and thus that said utterance has not yet begun; means for determining from said acoustic data the acoustic amplitude corresponding to successive portions of said data; means for producing an accumulated sound score which corresponds a sum of to the amount of said acoustic amplitude over said successive portions of data, said accumulated sound score inversely corresponding to the probability that an utterance has not yet begun; and means, responsive to both said silence score and said accumulated sound score for determining during which of said successive portions of data to look for the start of said utterance.

Claim 81

Dependent Claim 120 words

A speech recognition system as described in claim 69, wherein said means for determining during which successive portions of said data to look for the start of said utterance further includes: means for selecting during each successive portion of data the score from amoung said silence score and said accumulated sound score which indicates with the greatest probability that said utterance has not yet begun; and means for comparing said selected score with a threshold during each successive portion of data, and for causing said speech recognition system to look for the start of said utterance in the vicinity of said portion of data when said selected score for a given portion of said sequence is better than said threshold.

Claim 82

Dependent Claim 41 words

A speech recognition system as described in claim 69, wherein said means for producing an accumulated sound score includes means for accumulating the amount by which said acoustic amplitude exceeds an amplitude threshold during each of said successive portions of data.

Claim 83

Dependent Claim 150 words

A speech recognition method for recognizing a word in a sequence of acoustic data, said method comprising: storing an acoustic model of expected background noise; comprising said acoustic model with successive portions of said sequence of acoustic data and producing a silence score corresponding to the probability that such successive portions of data correspond to background noise and thus that said utterance has not yet begun; determining from said acoustic data the acoustic amplitude corresponding to successive portions of said data; producing an accumulated sound score which corresponds to a sum of the amount of said acoustic amplitude over said successive portions of data, said accumulated sound score inversely corresponding to the probability that an utterance has not yet begun; and determining, in response to both said silence score and said accumulated sound score, during which of said successive portions of data to look for the start of said utterance.

Claim 84

Dependent Claim 150 words

A speech recognition method for recognizing a word in a sequence of acoustic data, said method comprising: storing an acoustic model of expected background noise; comprising said acoustic model with successive portions of said sequence of acoustic data and producing a silence score corresponding to the probability that such successive portions of data correspond to background noise and thus that said utterance has not yet begun; determining from said acoustic data the acoustic amplitude corresponding to successive portions of said data; producing an accumulated sound score which corresponds to a sum of the amount of said acoustic amplitude over said successive portions of data, said accumulated sound score inversely corresponding to the probability that an utterance has not yet begun; and determining, in response to both said silence score and said accumulated sound score, during which of said successive portions of data to look for the start of said utterance.

Claim 85

Dependent Claim 114 words

A speech recognition method as described in claim 72, wherein said determining during which successive portions of said data to look for the start of said utterance further includes: selecting during each successive portion of data the score from amoung said silence score and said accumulated sound score which indicates with the greatest probability that said utterance has not yet begun; and comparing said selected score with a threshold during each successive portion of data, and for causing said speech recognition method to look for the start of said utterance in the vicinity of said portion of data when said selected score for a given portion of said sequence is better than said threshold.

Claim 86

Dependent Claim 38 words

A speech recognition method as described in claim 72, wherein said producing of an accumulated sound score includes accumulating the amount by which said acoustic amplitude exceeds an amplitude threshold during each of said successive portions of data.

Claim 87

Dependent Claim 314 words

A speech recognition system for recogizing a word in a sequence of acoustic data frames, said system comprising: means for storing an initial vocabulary of words; means for performing initial computations on said sequence of acoustic data frames and for producing, in response to said computations, an originally active vocabulary, containing those words from said initial vocabulary which are determined by said initial computations to be the most probable candidates for corresponding to said acoustic data frames; means for remembering the words contained in said originally active vocabulary; currently active vocabulary means for indicating which words in said originally active vocabulary belong to a currently active vocabulary; means for causing said currently active vocabulary means to initially indicate that all the words in said originally active vocabulary are in said currently active vocabulary; means for making a word match score corresponding to the probability that each word in said currently active vocabulary corresponds to said sequence of acoustic data frames, including means for making an estimation cycle for each successive data frame, in which for each word in said currently active vocabulary, a comparison is made between the acoustic model of said word and said frame and said word match score for said word is updated in response to said comparison; means for causing said currently active vocabulary means to indicate that a given word is no longer in said currently active vocabulary when the word match score for said word becomes worse than a given threshold; means for producing a start score corresponding to the probability that an acoustic data frame precedes the beginning of the utterance of the word to be recognized; and means for resetting said currently active vocabulary to include all the words remembered by said means for remembering as belonging in said original vocabulary and for performing said resetting when said start score is better than a given threshold.

Claim 88

Dependent Claim 314 words

A speech recognition system for recogizing a word in a sequence of acoustic data frames, said system comprising: means for storing an initial vocabulary of words; means for performing initial computations on said sequence of acoustic data frames and for producing, in response to said computations, an originally active vocabulary, containing those words from said initial vocabulary which are determined by said initial computations to be the most probable candidates for corresponding to said acoustic data frames; means for remembering the words contained in said originally active vocabulary; currently active vocabulary means for indicating which words in said originally active vocabulary belong to a currently active vocabulary; means for causing said currently active vocabulary means to initially indicate that all the words in said originally active vocabulary are in said currently active vocabulary; means for making a word match score corresponding to the probability that each word in said currently active vocabulary corresponds to said sequence of acoustic data frames, including means for making an estimation cycle for each successive data frame, in which for each word in said currently active vocabulary, a comparison is made between the acoustic model of said word and said frame and said word match score for said word is updated in response to said comparison; means for causing said currently active vocabulary means to indicate that a given word is no longer in said currently active vocabulary when the word match score for said word becomes worse than a given threshold; means for producing a start score corresponding to the probability that an acoustic data frame precedes the beginning of the utterance of the word to be recognized; and means for resetting said currently active vocabulary to include all the words remembered by said means for remembering as belonging in said original vocabulary and for performing said resetting when said start score is better than a given threshold.

Claim 89

Dependent Claim 104 words

A speech recognition system as described in claim 75, wherein: said currently active vocabulary means includes means for keeping all the words in the currently active vocabulary as a linked list; said means for causing said currently active vocabulary means to indicate that a given word is no longer in said currently active vocabulary includes means for causing said given word to be unlinked from said linked list; and said means for resetting said currently active vocabulary includes means for causing all the words remembered by said means for remembering as being in said originally active vocabulary to be relinked into said linked list.

Claim 90

Dependent Claim 126 words

A speech recognition system as described in claim 76, further including means for representing said initial vocabulary as a list of word records, each of which has a means for storing a pointer to another word record in said list; and said means for causing said currently active vocabulary means to originally indicate that all the words in said originally active vocabulary are in said currently active vocabulary includes means for causing said currently active vocabulary means to point to said word record representing a first word in said originally active vocabulary and to cause the pointer in the word record associated with each word in a said originally active vocabulary to point to the word record of the next word in said originally active vocabulary.

Claim 91

Dependent Claim 275 words

A speech recognition method for recognizing a word in a sequence of acoustic data frames, said method comprising: storing an initial vocabulary of words; performing initial computations on said sequence of acoustic data frames and for producing, in response to said computations, an originally active vocabulary, containing those words from said initial vocabulary which are determined by said initial computations to be the most probable candidates for corresponding to said acoustic data frames; remembering the words contained in said originally active vocabulary; indicating which words in said originally active vocabulary belong to a currently active vocabulary; initially indicating that all the words in said originally active vocabulary are in said currently active vocabulary; making a word match score corresponding to the probability that each word in said currently active vocabulary corresponds to said sequence of acoustic data frames, including making an estimation cycle for each successive data frame, in which for each word in said currently active vocabulary, a comparison is made between the acoustic model of said word and said frame and said word match score for said word is updated in response to said comparison; indicating that a given word is no longer in said currently active vocabulary when the word match score for said word becomes worse than a given threshold; producing a start score corresponding to the probability that an acoustic data frame precedes the beginning of the utterance of the word to be recognized; and resetting said currently active vocabulary to include all the words remembered by said remembering as belonging in said original vocabulary and for performing said resetting when said start score is better than a given threshold.

Claim 92

Dependent Claim 275 words

A speech recognition method for recognizing a word in a sequence of acoustic data frames, said method comprising: storing an initial vocabulary of words; performing initial computations on said sequence of acoustic data frames and for producing, in response to said computations, an originally active vocabulary, containing those words from said initial vocabulary which are determined by said initial computations to be the most probable candidates for corresponding to said acoustic data frames; remembering the words contained in said originally active vocabulary; indicating which words in said originally active vocabulary belong to a currently active vocabulary; initially indicating that all the words in said originally active vocabulary are in said currently active vocabulary; making a word match score corresponding to the probability that each word in said currently active vocabulary corresponds to said sequence of acoustic data frames, including making an estimation cycle for each successive data frame, in which for each word in said currently active vocabulary, a comparison is made between the acoustic model of said word and said frame and said word match score for said word is updated in response to said comparison; indicating that a given word is no longer in said currently active vocabulary when the word match score for said word becomes worse than a given threshold; producing a start score corresponding to the probability that an acoustic data frame precedes the beginning of the utterance of the word to be recognized; and resetting said currently active vocabulary to include all the words remembered by said remembering as belonging in said original vocabulary and for performing said resetting when said start score is better than a given threshold.

Claim 93

Dependent Claim 91 words

A speech recognition method as described in claim 78, wherein: said indicating which words belong to a currently active vocabulary includes keeping all the words in the currently active vocabulary as a linked list; said indicating that a given word is no longer in said currently active vocabulary includes causing said given word to be unlinked from said linked list; and said resetting of said currently active vocabulary includes causing all the words remembered by said remembering as being in said originally active vocabulary to be relinked into said linked list.

Claim 94

Dependent Claim 116 words

A speech recognition method as described in claim 79, further including representing said initial vocabulary as a list of word records, each of which has a pointer capable of pointing to another word record in said list; and said originally indicating that all the words in said originally active vocabulary are in said currently active vocabulary includes causing a pointer to point to said word record representing a first word in said originally active vocabulary and to cause the pointers in the word record associated with each word in a said originally active vocabulary to point to the word record of the next word in said originally active vocabulary when there is such a next word.

Claim 95

Dependent Claim 336 words

A likelihood processor for use in a speech recognition system to compute a score corresponding to the probabilistic match between a vector of parameters associated with a frame of acoustic data and a node model having for each parameter in said vector both an expected value and a weighting value corresponding to the expected deviation from said expected value, said processor comprising: parameter supplying means for supplying a digital value corresponding to a given parameter of a given frame of data; expected value supplying means for supplying a digital value corresponding to the expected value for said given parameter according to a given node model; first hardware means, receiving inputs from both said parameter supplying means and said expected value supplying means, for computing the absolute value of the difference between said two inputs and for producing said absolute value as a digital output; weighting value supplying means for supplying a digital weighting value corresponding to the expected deviation from said expected value for said given parameter according to said given node model; memory means, receiving addressing inputs from both the output of said first hardware means and said weighting value supplying means, for storing, at each of a plurality of addresses formed by combining said two addressing inputs, a parameter match score corresponding to a probability that said given parameter value corresponds to said expected value, considering said weighting value, and for producing said score as a digital output; a hardware adder having two inputs for receiving digital values and an output at which it produces a digital value corresponding to the sum of the values supplied to its inputs; means for connecting the output of said memory means to one input of said adder; latch means connected to the output of said adder for latching the value produced at said output; and means for supplying the value latched by said latch means to one of the inputs of said adder, whereby said hardware adder can produce a running total of said parameter match scores.

Claim 96

Dependent Claim 336 words

A likelihood processor for use in a speech recognition system to compute a score corresponding to the probabilistic match between a vector of parameters associated with a frame of acoustic data and a node model having for each parameter in said vector both an expected value and a weighting value corresponding to the expected deviation from said expected value, said processor comprising: parameter supplying means for supplying a digital value corresponding to a given parameter of a given frame of data; expected value supplying means for supplying a digital value corresponding to the expected value for said given parameter according to a given node model; first hardware means, receiving inputs from both said parameter supplying means and said expected value supplying means, for computing the absolute value of the difference between said two inputs and for producing said absolute value as a digital output; weighting value supplying means for supplying a digital weighting value corresponding to the expected deviation from said expected value for said given parameter according to said given node model; memory means, receiving addressing inputs from both the output of said first hardware means and said weighting value supplying means, for storing, at each of a plurality of addresses formed by combining said two addressing inputs, a parameter match score corresponding to a probability that said given parameter value corresponds to said expected value, considering said weighting value, and for producing said score as a digital output; a hardware adder having two inputs for receiving digital values and an output at which it produces a digital value corresponding to the sum of the values supplied to its inputs; means for connecting the output of said memory means to one input of said adder; latch means connected to the output of said adder for latching the value produced at said output; and means for supplying the value latched by said latch means to one of the inputs of said adder, whereby said hardware adder can produce a running total of said parameter match scores.

Claim 97

Dependent Claim 76 words

A likelihood processor as described in claim 81, wherein said first hardware means includes means for limiting the numerical value of its output to a limited range which is less than the range of absolute values made possible by the range of values of its inputs, whereby the memory space required in said memory means to correspond to the address space formed by combining the output of said hardware means and said weighting value is reduced.

Claim 98

Dependent Claim 67 words

A likelihood processor as described in claim 82, wherein said means for limiting the numerical value of the output of said first hardware means includes means for causing the numerical value of said output to be equal to a value at the high end of said limited range when the actual absolute value of the difference between the inputs of said hardware means exceeds said limited range.

Claim 99

Dependent Claim 64 words

A likelihood processor as described in claim 81, wherein said memory means stores, at each of said addresses, a parameter match score which corresponds to the sum of two portions, a first portion which corresponds to the product of the output of said first hardware means and said digital weighting value and a second portion which corresponds to the logarithm of said weighting value.

Claim 100

Dependent Claim 303 words

A likelihood processor for use in a speech recognition system to compute a score corresponding to the probabilistic match between a vector of parameters associated with a frame of acoustic data and a node model having for each parameter in said vector both an expected value and a weighting value corresponding to the expected deviation from said expected value, said processor comprising: parameter supplying means for supplying a digital value corresponding to a given parameter of a given frame of data; expected value supplying means for supplying a digital value corresponding to the expected value for said given parameter according to a given node model; subtracting means, receiving inputs from both said parameter supplying means and said expected value supplying means, for computing the absolute value of the difference between said two inputs and for producing said absolute value as a digital output; weighting value supplying means for supplying a digital weighting value corresponding to the expected deviation from said expected value for said given parameter according to said given node model; memory means, receiving addressing inputs from both the output of said subtracting means and said weighting value supplying means, for storing, at each of a plurality of addresses formed by combining said addressing inputs, a parameter match score corresponding to a probability that said given parameter value corresponds to said expected value given said weighting value and for producing said score as a digital output, said parameter match score having a value which corresponds to the sum of two portions, a first portion which corresponds to the product of the output of said subtracting means and said digital weighting value and a second portion which corresponds to the logarithm of said weighting value; and accumulator means for accumulating said parameter match scores associated with a given frame for a given node.

Claim 101

Dependent Claim 303 words

A likelihood processor for use in a speech recognition system to compute a score corresponding to the probabilistic match between a vector of parameters associated with a frame of acoustic data and a node model having for each parameter in said vector both an expected value and a weighting value corresponding to the expected deviation from said expected value, said processor comprising: parameter supplying means for supplying a digital value corresponding to a given parameter of a given frame of data; expected value supplying means for supplying a digital value corresponding to the expected value for said given parameter according to a given node model; subtracting means, receiving inputs from both said parameter supplying means and said expected value supplying means, for computing the absolute value of the difference between said two inputs and for producing said absolute value as a digital output; weighting value supplying means for supplying a digital weighting value corresponding to the expected deviation from said expected value for said given parameter according to said given node model; memory means, receiving addressing inputs from both the output of said subtracting means and said weighting value supplying means, for storing, at each of a plurality of addresses formed by combining said addressing inputs, a parameter match score corresponding to a probability that said given parameter value corresponds to said expected value given said weighting value and for producing said score as a digital output, said parameter match score having a value which corresponds to the sum of two portions, a first portion which corresponds to the product of the output of said subtracting means and said digital weighting value and a second portion which corresponds to the logarithm of said weighting value; and accumulator means for accumulating said parameter match scores associated with a given frame for a given node.

Claim 102

Dependent Claim 329 words

A likelihood processor for use in a speech recognition system to compute a score corresponding to the probabilistic match between a vector of parameters associated with a frame of acoustic data and a node model having for each parameter in said vector both an expected value for said parameter and a weighting value corresponding to the expected deviation from said expected value, said processor comprising: parameter supplying means for supplying a digital value corresponding to a given parameter of a given frame of data; expected value supplying means for supplying a digital value corresponding to the expected value for said given parameter according to a given node model; subtracting means, receiving inputs from both said parameter supplying means and said expected value supplying means, for computing the absolute value of the difference between said two inputs and for producing said absolute value as a digital output; weighting value supplying means for supplying a digital weighting value corresponding to the expected deviation from said expected value for said given parameter according to said given node model; memory means, receiving addressing inputs from both the output of said subtracting means and said weighting value supplying means, for storing, at each of a plurality of addresses formed by combining said addressing inputs, a parameter match score having a value corresponding to a probability that said given parameter value corresponds to said expected value given said weighting value and for producing said match score as a digital output; and accumulator means for accumulating said parameter match scores associated with a given frame for a given node; said subtracting means including means for limiting the numerical value of its output to a limited range which is less than the range of the absolute values made possible by the range of values of its inputs, whereby the memory space required in said memory means to correspond to the address space formed by combining the output of said subtacting means and said weighting value is reduced.

Claim 103

Dependent Claim 329 words

A likelihood processor for use in a speech recognition system to compute a score corresponding to the probabilistic match between a vector of parameters associated with a frame of acoustic data and a node model having for each parameter in said vector both an expected value for said parameter and a weighting value corresponding to the expected deviation from said expected value, said processor comprising: parameter supplying means for supplying a digital value corresponding to a given parameter of a given frame of data; expected value supplying means for supplying a digital value corresponding to the expected value for said given parameter according to a given node model; subtracting means, receiving inputs from both said parameter supplying means and said expected value supplying means, for computing the absolute value of the difference between said two inputs and for producing said absolute value as a digital output; weighting value supplying means for supplying a digital weighting value corresponding to the expected deviation from said expected value for said given parameter according to said given node model; memory means, receiving addressing inputs from both the output of said subtracting means and said weighting value supplying means, for storing, at each of a plurality of addresses formed by combining said addressing inputs, a parameter match score having a value corresponding to a probability that said given parameter value corresponds to said expected value given said weighting value and for producing said match score as a digital output; and accumulator means for accumulating said parameter match scores associated with a given frame for a given node; said subtracting means including means for limiting the numerical value of its output to a limited range which is less than the range of the absolute values made possible by the range of values of its inputs, whereby the memory space required in said memory means to correspond to the address space formed by combining the output of said subtacting means and said weighting value is reduced.

Claim 104

Dependent Claim 66 words

A likelihood processor as described in claim 86, wherein said means for limiting the numerical value of the output of said subtracting means includes means for causing the numerical value of said output to be equal to a value at the high end of said limited range when the actual absolute value of the difference between the inputs of said subtracting means exceeds said limited range.

Claim 105

Dependent Claim 68 words

A speech recognition system for recognizing a sequence of separately spoken words, said system comprising: means for responding to a given utterance by producing a best guess word which is the word considered by the system most likely to correspond to said utterance; and means for responding to the utterance of another word as confirmation that said best guess word is the word corresponding to said given utterance.

Claim 106

Dependent Claim 68 words

A speech recognition system for recognizing a sequence of separately spoken words, said system comprising: means for responding to a given utterance by producing a best guess word which is the word considered by the system most likely to correspond to said utterance; and means for responding to the utterance of another word as confirmation that said best guess word is the word corresponding to said given utterance.

Claim 107

Dependent Claim 61 words

A speech recognition system as described in claim 88, wherein said means for responding to the utterance of another word as confirmation includes means for detecting an utterance and means for responding to any detection of another utterance by said means as confirmation that said best guess word is the word corresponding to said given utterance spoken before said another utterance.

Claim 108

Dependent Claim 28 words

A speech recognition system as described in claim 88, further including means for displaying said best guess word so a user can see it before uttering another word.

Claim 109

Dependent Claim 97 words

A speech recognition system as described in claim 90: wherein said means for responding to an utterance includes means for producing in response to said given utterance one or more close call words which are considered by the system next most likely to correspond to said given utterance after said best guess word; further including means for displaying said one or more close call words; and further including means for responding to an input signal associated with a given close call word as confirmation that said given close word is the word corresponding to said given utterance.

Claim 110

Dependent Claim 36 words

A speech recognition system as described in claim 91, wherein said means for displaying one or more close call words further includes means for indicating which input signal is associated with said given close call word.

Claim 111

Dependent Claim 33 words

A speech recognition system as described in claim 91, wherein said means for responding to an input signal includes means for responding to an actuation of a keyboard key as said input signal.

Claim 112

Dependent Claim 58 words

A speech recognition system as described in claim 93, further including means for responding to an input signal associated with the rejection of said best guess word and said one or more close call words as confirmation that neither said best guess word nor any of said close call words was the word corresponding to said given utterance.

Claim 113

Dependent Claim 64 words

A speech recognition system as described in claim 94, wherein said means for responding to an input signal associated with the rejection of said best guess word and said one or more close call words includes means for removing the display of said best guess word and said one or more close call words in response to said input signal associated with said rejection.

Claim 114

Dependent Claim 53 words

A speech recognition system as described in claim 94, wherein said means for responding to an input signal associated with the rejection of said best guess word and said one or more close call words includes means for responding to an actuation of a keyboard key as said input signal associated with rejection.

Claim 115

Dependent Claim 45 words

A speech recognition system as described in claim 91, wherein said means for responding to the utterance of another word as confirmation that said best guess word was the intended word includes means for terminating the displaying of said one or more close call words.

Claim 116

Dependent Claim 113 words

A speech recognition system as described in claim 88, further including a language model means for helping it predict the likelihood that a certain utterance corresponds to one or more certain words based at least in part on the previous word recognized by said system, wherein said means for responding to the utterance of another word as confirmation that said best guess word is the word corresponding to said given utterance includes means for indicating to said language model means that said best guess word is the previous word for use of said language model means in helping to predict the likelihood that said another utterance corresponds to one or more certain words.

Claim 117

Dependent Claim 158 words

A speech recognition system as described in claim 98: wherein said means for responding to a given utterance includes means for producing in response to said given utterance one or more close call words which are considered by the system next most likely, after said best guess word, to correspond to said given utterance; further including means for displaying said one or more close call words; further including means for responding to an input signal associated with a given close call word as confirmation that said given close call word is the word corresponding to said given utterance; and wherein said means for responding to an input signal associated with a given close call word includes means for indicating to said language model means that said given close call word is the previous word for use of said language model means in helping to predict the likelihood that said another utterance corresponds to one or more certain words.

Claim 118

Dependent Claim 63 words

A speech recognition method for recognizing a sequence of spoken words, said method comprising: responding to a given utterance by producing a best guess word which is the word considered by the method most likely to correspond to said utterance; and responding to the utterance of another word as confirmation that said best guess word is the word corresponding to said given utterance.

Claim 119

Dependent Claim 63 words

A speech recognition method for recognizing a sequence of spoken words, said method comprising: responding to a given utterance by producing a best guess word which is the word considered by the method most likely to correspond to said utterance; and responding to the utterance of another word as confirmation that said best guess word is the word corresponding to said given utterance.

Claim 120

Dependent Claim 55 words

A speech recognition method as described in claim 100, wherein said responding to the utterance of another word as confirmation includes detecting an utterance and responding to any detection of another utterance by said means as confirmation that said best guess word is the word corresponding to said given utterance spoken before said another utterance.

Claim 121

Dependent Claim 26 words

A speech recognition method as described in claim 100, further including displaying said best guess word so a user can see it before uttering another word.

Claim 122

Dependent Claim 89 words

A speech recognition method as described in claim 102: wherein said responding to an utterance includes producing in response to said given utterance one or more close call words which are considered by the method next most likely to correspond to said given utterance after said best guess word; further including displaying said one or more close call words; and further including responding to an input signal associated with a given close call word as confirmation that said given close word is the word corresponding to said given utterance.

Claim 123

Dependent Claim 32 words

A speech recognition method as described in claim 103, wherein said displaying one or more close call words further includes indicating which input signal is associated with said given close call word.

Claim 124

Dependent Claim 29 words

A speech recognition method as described in claim 103, wherein said responding to an input signal includes responding to an actuation of a keyboard key as said input signal.

Claim 125

Dependent Claim 56 words

A speech recognition method as described in claim 105, further including responding to an input signal associated with the rejection of said best guess word and said one or more close call words as confirmation that neither said best guess word nor any of said close call words was the word corresponding to said given utterance.

Claim 126

Dependent Claim 60 words

A speech recognition method as described in claim 106, wherein said responding to an input signal associated with the rejection of said best guess word and said one or more close call words includes removing the display of said best guess word and said one or more close call words in response to said input signal associated with said rejection.

Claim 127

Dependent Claim 49 words

A speech recognition method as described in claim 106, wherein said responding to an input signal associated with the rejection of said best guess word and said one or more close call words includes responding to an actuation of a keyboard key as said input signal associated with rejection.

Claim 128

Dependent Claim 41 words

A speech recognition method as described in claim 103, wherein said responding to the utterance of another word as confirmation that said best guess word was the intended word includes terminating the displaying of said one or more close call words.

Claim 129

Dependent Claim 106 words

A speech recognition method as described in claim 100, further including using a language model to help in the prediction of a likelihood that a certain utterance corresponds to one or more certain words based at least in part on the previous word recognized by said method, wherein said responding to the utterance of another word as confirmation that said best guess word is the word corresponding to said given utterance includes indicating that said best guess word is the previous word for said use of said language model in helping to predict the likelihood that said another utterance corresponds to one or more certain words.

Claim 130

Dependent Claim 141 words

A speech recognition method as described in claim 110: wherein said responding to a given utterance includes producing in response to said given utterance one or more close call words which are considered by the method next most likely, after said best guess word, to correspond to said given utterance; further including displaying said one or more close call words; further including responding to an input signal associated with a given close call word as confirmation that said given close call word is the word corresponding to said given utterance; and wherein said responding to an input signal associated with a given close call word includes indicating that said given close call word is the previous word for said use of said language model in helping to predict the likelihood that said another utterance corresponds to one or more certain words.

Claim 131

Dependent Claim 141 words

A speech recognition system for recognizing a sequence of speech sounds, said system comprising: means for responding to said sequence of speech sounds by producing a best guess word, which is the word considered by the system most likely to correspond to said sequence, and one or more close call words, which are considered by the system next most likely to correspond to said given sequence after said best guess word; and means for displaying said best guess word so a user can see it; means for displaying said one or more close call words so a user can see it; and means for responding to an input signal associated with a given close call word as confirmation that said given close call word is the word corresponding to said given sequence of speech sounds, rather than said best guess word.

Claim 132

Dependent Claim 141 words

A speech recognition system for recognizing a sequence of speech sounds, said system comprising: means for responding to said sequence of speech sounds by producing a best guess word, which is the word considered by the system most likely to correspond to said sequence, and one or more close call words, which are considered by the system next most likely to correspond to said given sequence after said best guess word; and means for displaying said best guess word so a user can see it; means for displaying said one or more close call words so a user can see it; and means for responding to an input signal associated with a given close call word as confirmation that said given close call word is the word corresponding to said given sequence of speech sounds, rather than said best guess word.

Claim 133

Dependent Claim 93 words

A speech recognition system as described in claim 112, wherein said means for displaying said best guess word includes means for displaying said word in the context of words spoken before or after said sequence of speech sounds and said means for displaying said one or more close call words further includes means for displaying said one or more close call words in close proximity to said best guess word, so that said close call words are also displayed in the context of words spoken before or after a sequence of speech sounds.

Claim 134

Dependent Claim 38 words

A speech recognition system as described in claim 112, wherein said means for displaying one or more close call words further includes means for displaying an indication which input signal is associated with said given close call word.

Claim 135

Dependent Claim 33 words

A speech recognition system as described in claim 112, wherein said means for responding to an input signal includes means for responding to an actuation of a keyboard key as said input signal.

Claim 136

Dependent Claim 62 words

A speech recognition system as described in claim 112, further including means for responding to an input signal associated with the rejection of both said best guess word and said one or more close call words as confirmation that neither said best guess word nor any of said close call words was the word corresponding to said given sequence of speech sounds.

Claim 137

Dependent Claim 65 words

A speech recognition system as described in claim 116, wherein said means for responding to an input signal associated with the rejection of both said best guess word and said one or more close call words includes means for removing the display of said best guess word and said one or more close call words in response to said input signal associated with said rejection.

Claim 138

Dependent Claim 54 words

A speech recognition system as described in claim 117, wherein said means for responding to an input signal associated with the rejection of both said best guess word and said one or more close call words includes means for responding to an actuation of a keyboard key as said input signal associated with rejection.

Claim 139

Dependent Claim 104 words

A speech recognition system as described in claim 112: further including means for using a language model to help in the prediction of a likelihood that a certain utterance corresponds to one or more certain words based at least in part on the previous word recognized by said system; and wherein said means for responding to an input signal associated with a given close call word includes means for indicating that said given close call word is the previous word for said use of said language model in helping to predict the likelihood that said another utterance corresponds to one or more certain words.

Claim 140

Dependent Claim 133 words

A speech recognition method for recognizing a sequence of speech sounds, said method comprising: responding to said sequence of speech sounds by producing a best guess word, which is the word considered by the method most likely to correspond to said sequence, and one or more close call words, which are considered by the method next most likely to correspond to said given sequence after said best guess word; and displaying said best guess word so a user can see it; displaying said one or more close call words so a user can see it; and responding to an input signal associated with a given close call word as confirmation that said given close call word is the word corresponding to said given sequence of speech sounds, rather than said best guess word.

Claim 141

Dependent Claim 133 words

A speech recognition method for recognizing a sequence of speech sounds, said method comprising: responding to said sequence of speech sounds by producing a best guess word, which is the word considered by the method most likely to correspond to said sequence, and one or more close call words, which are considered by the method next most likely to correspond to said given sequence after said best guess word; and displaying said best guess word so a user can see it; displaying said one or more close call words so a user can see it; and responding to an input signal associated with a given close call word as confirmation that said given close call word is the word corresponding to said given sequence of speech sounds, rather than said best guess word.

Claim 142

Dependent Claim 87 words

A speech recognition method as described in claim 120, wherein said displaying of said best guess word includes displaying said word in the context of words spoken before or after said sequence of speech sounds and said displaying of said one or more close call words further includes displaying said one or more close call words in close proximity to said best guess word, so that said close call words are also displayed in the context of words spoken before or after a sequence of speech sounds.

Claim 143

Dependent Claim 35 words

A speech recognition method as described in claim 120, wherein said displaying of one or more close call words further includes displaying an indication which input signal is associated with said given close call word.

Claim 144

Dependent Claim 29 words

A speech recognition method as described in claim 120, wherein said responding to an input signal includes responding to an actuation of a keyboard key as said input signal.

Claim 145

Dependent Claim 60 words

A speech recognition method as described in claim 120, further including responding to an input signal associated with the rejection of both said best guess word and said one or more close call words as confirmation that neither said best guess word nor any of said close call words was the word corresponding to said given sequence of speech sounds.

Claim 146

Dependent Claim 61 words

A speech recognition method as described in claim 124, wherein said responding to an input signal associated with the rejection of both said best guess word and said one or more close call words includes removing the display of said best guess word and said one or more close call words in response to said input signal associated with said rejection.

Claim 147

Dependent Claim 50 words

A speech recognition method as described in claim 125, wherein said responding to an input signal associated with the rejection of both said best guess word and said one or more close call words includes responding to an actuation of a keyboard key as said input signal associated with rejection.

Claim 148

Dependent Claim 98 words

A speech recognition method as described in claim 120: further including using a language model to help in the prediction of a likelihood that a certain utterance corresponds to one or more certain words based at least in part on the previous word recognized by said method; and wherein said responding to an input signal associated with a given close call word includes indicating that said given close call word is the previous word for said use of said language model in helping to predict the likelihood that said another utterance corresponds to one or more certain words.

Annuity Stage	Window Opens	Payment Deadline	Surcharge Period Until	Estimated Fee	Status
3.5 Years (Year 4)	1991-11-08	1992-05-09	1992-11-08	$2000.00	Window Closed / Lapsed
7.5 Years (Year 8)	1995-11-08	1996-05-09	1996-11-08	$3760.00	Window Closed / Lapsed
11.5 Years (Year 12)	1999-11-08	2000-05-09	2000-11-08	$7700.00	Window Closed / Lapsed

Speech recognition apparatus and method

Abstract

Inventors

Bibliographic Details

Claims Analysis

References Cited (Backward Citations)

USPTO Annuity & Maintenance Windows

Assigned Categories & Tags

Assign Category Tag