Large-vocabulary continuous speech prefiltering and processing system

Claim 3

A prefiltering system for processing continuous speech, said speech including a succession of utterances, comprising: A. cluster data storage means for storing a plurality of M cluster data sets, C 1 , . . . , C M , where M is an integer greater than 1, each of said cluster data sets including data representative of a plurality of word models; B. frame data means for generating succession of w frame data sets v t , v t +1, . . . v t +w-1, beginning at a frame start time t during said succession of utterances, where w is an integer greater than 1, said succession of frame data sets being representative of a corresponding succession of temporal segments of said utterances, each of said frame data sets including k values representative of different frame parameters, where k≧1; C. data reduction means selectively operable on said w frame data sets for generating s reduced frame data sets where s<w, each of said reduced frame data sets Y 1 , Y 2 , . . . , Y 3 , being related to an associated plurality of said frame data sets and including j values representative of different reduced frame data set parameters; D. scoring means for evaluating each of said reduced frame data sets against a succession of said cluster data sets to generate a cluster score S Y for each of said cluster data sets; E. selectively operable identifying means for identifying each of said word models of said cluster data sets having a cluster score bearing a predetermined relation to at least one threshold score T, said identified word models defining a candidate word list; F. control means for determining said frame start times t, where successive start times t are spaced apart arbitrarily, said frame start times being independent of identification of an initial anchor; and G. means for generating a signal representative of said candidate word list for preselected ones of said frame start times t determined by said control means.

Claim 4

A prefiltering system for processing continuous speech, said speech including a succession of utterances, comprising: A. cluster data storage means for storing a plurality of M cluster data sets, C 1 , . . . , C M , where M is an integer greater than 1, each of said cluster data sets including data representative of a plurality of word models; B. frame data means for generating succession of w frame data sets v t , v t +1, . . . v t +w-1, beginning at a frame start time t during said succession of utterances, where w is an integer greater than 1, said succession of frame data sets being representative of a corresponding succession of temporal segments of said utterances, each of said frame data sets including k values representative of different frame parameters, where k≧1; C. data reduction means selectively operable on said w frame data sets for generating s reduced frame data sets where s<w, each of said reduced frame data sets Y 1 , Y 2 , . . . , Y 3 , being related to an associated plurality of said frame data sets and including j values representative of different reduced frame data set parameters; D. scoring means for evaluating each of said reduced frame data sets against a succession of said cluster data sets to generate a cluster score S Y for each of said cluster data sets; E. selectively operable identifying means for identifying each of said word models of said cluster data sets having a cluster score bearing a predetermined relation to at least one threshold score T, said identified word models defining a candidate word list; F. control means for determining said frame start times t, where successive start times t are spaced apart arbitrarily, said frame start times being independent of identification of an initial anchor; and G. means for generating a signal representative of said candidate word list for preselected ones of said frame start times t determined by said control means.

Claim 5

A system according to claim 2 wherein said cluster data storage means and said frame data means are adapted whereby said succession of frame data sets is associated with duration D 1 , and wherein said cluster data sets are each associated with duration≦D 2 , such that: D.sub.1 ≧D.sub.2.

Claim 6

A system according to claim 2 wherein said cluster data storage means is adapted whereby said cluster data sets are wordstart cluster data sets.

Claim 7

A system according to claim 2 wherein said cluster data storage means is adapted whereby said word models of each of said cluster data sets correspond to acoustically similar utterances over a succession of no more than w frame data sets.

Claim 8

A system according to claim 2 wherein said data reduction means includes smooth frame means for processing said frame data sets whereby said reduced frame data sets are smoothed frame data sets.

Claim 9

A system according to claim 6 wherein said smooth frame means is adapted for determining smooth frame data sets in accordance with: ##EQU4## where Y 1 , Y 2 . . . Y s are said smooth frame data sets, and whereby each of said smooth frame data sets is associated with b of said frame data sets, and wherein said b are integers greater than 1, a i are predetermined weighting coefficients, and c defines an offset of each smooth frame data set with respect to the next previous smooth frame data set.

Claim 10

A system according to claim 7 wherein said data reduction means is adapted whereby w=b+(s-1)c.

Claim 11

A system according to claim 8 wherein said data reduction means is adapted whereby w=12.

Claim 12

A system according to claim 8 wherein said data reduction means is adapted whereby k=8.

Claim 13

A system according to claim 10 wherein said data reduction means is adapted whereby s=3.

Claim 14

A system according to claim 11 wherein said data reduction means is adapted whereby b=4.

Claim 15

A system according to claim 11 wherein said data reduction means is adapted whereby c=4.

Claim 16

A system according to claim 8 wherein w=12, k=8, s=3, b=4, c=4, and a i =1/b.

Claim 17

A system according to claim 7 wherein said data reduction means is adapted whereby said weighting coefficients a i =l/b.

Claim 18

A system according to claim 2 wherein said cluster data storage means is adapted whereby each of said word models includes r node data vectors f 1 , . . . , f r , where r≦s, each of said node data vectors being representative of a characteristic related to the occurrence of a selected one of acoustic segments from each word of a set of words associated with said cluster data sets in said acoustic segment of said utterances.

Claim 19

A system according to claim 16 wherein said cluster data storage means is adapted whereby said characteristic related to the occurrence of a selected one of acoustic segments comprises a probability distribution.

Claim 20

A system according to claim 17 wherein said cluster data storage means is adapted whereby said cluster score S Y corresponds to: ##EQU5## wherein Y i are ones of Y 1 , Y 2 , . . . , Y s , and r are said node data vectors f 1 , . . . , f r for corresponding ones of said word models.

Claim 21

A system according to claim 2 wherein said data reduction means is adapted whereby j=k.

Claim 22

A system according to claim 2 wherein indentifying means includes a first identifying means for identifying each of said cluster data sets having a cluster score measured with respect to a first predetermined threshold T 1 , said identified cluster data sets defining preliminary cluster data sets.

Claim 23

A system according to claim 20 further comprising means for generating a word score representative of the sum of said cluster score S Y of said cluster data set associated with each of said word models and a language model score S L and wherein said identifying means further comprises a selectively operable second identifying means for identifying each of said word models of each of said cluster data sets having a word score S Y measured with respect to a second threshold T 2 , said sum represented by: S.sub.W =S.sub.Y +S.sub.L.

Claim 24

A continuous speech processing method for processing speech including a succession of utterances comprising the steps of: A. storing a plurality of M cluster data sets, C 1 , . . . , C M , where M is an integer greater than 1, each of said cluster data sets including data representative of a plurality of word models; B. generating a succession of w frame data sets v t , v t +1, . . . v t +w-1, beginning at a frame start time t during said succession of utterances, where w is an integer greater than 1, each of said frame data sets being representative of successive acoustic segments of utterances for a frame period, each of said frame data sets including k values representative of different frame parameters where k≧1; C. reducing w of said frame data sets to generate s reduced frame data sets Y 1 , Y 2 , . . . , Y 3 , where s<w, each of said reduced frame data sets being related to an associated plurality of said frame data sets and including j related to the k values of said associated frame data sets, where j≦k; D. evaluating said reduced frame data sets with a succession of said cluster data sets to generate a cluster score S Y for each of said cluster data sets; E. identifying each of said word models having a cluster score bearing a predetermined relation to at least one threshold score T, said identified word models defining a word list; F. determining said frame start times, t, where successive start times t are identified at arbitrarily selected intervals, said frame start times being independent of identification of an initial anchor; and G. generating a signal representative of said candidate word list for selected ones of said frame start times determined in step F.

Claim 25

A continuous speech processing method for processing speech including a succession of utterances comprising the steps of: A. storing a plurality of M cluster data sets, C 1 , . . . , C M , where M is an integer greater than 1, each of said cluster data sets including data representative of a plurality of word models; B. generating a succession of w frame data sets v t , v t +1, . . . v t +w-1, beginning at a frame start time t during said succession of utterances, where w is an integer greater than 1, each of said frame data sets being representative of successive acoustic segments of utterances for a frame period, each of said frame data sets including k values representative of different frame parameters where k≧1; C. reducing w of said frame data sets to generate s reduced frame data sets Y 1 , Y 2 , . . . , Y 3 , where s<w, each of said reduced frame data sets being related to an associated plurality of said frame data sets and including j related to the k values of said associated frame data sets, where j≦k; D. evaluating said reduced frame data sets with a succession of said cluster data sets to generate a cluster score S Y for each of said cluster data sets; E. identifying each of said word models having a cluster score bearing a predetermined relation to at least one threshold score T, said identified word models defining a word list; F. determining said frame start times, t, where successive start times t are identified at arbitrarily selected intervals, said frame start times being independent of identification of an initial anchor; and G. generating a signal representative of said candidate word list for selected ones of said frame start times determined in step F.

Claim 26

A method according to claim 22 wherein said reducing step C further comprises the substep of smoothing said frame data sets.

Claim 27

A method according to claim 23 wherein said smoothing substep includes smoothing in accordance with ##EQU6## wherein Y 1 , Y 2 . . . Y s are said smooth frame data sets, and wherein said b are integers greater than 1, a i are predetermined weighting coefficients, and c defines an offset of each smooth frame data set with respect to the next previous smooth frame data set.

Claim 28

A method according to claim 24 wherein said smoothing substep includes setting w=12, k=8, s=3, b=4, c=4, and a i =1/b.

Claim 29

A method according to claim 22 wherein said storing step includes the substep of generating for each of said word models, r node data vectors f 1 , . . . , f r , where r is less than or equal to s, each of said node data vectors being representative of a characteristic related to the occurrence of a selected one of acoustic segments from each word of a set of words associated with said cluster data sets in said acoustic segment of said utterances, and wherein said evaluating step includes the substep of determining said cluster score S Y in accordance with: ##EQU7## wherein Y i are ones of Y 1 , Y 2 , . . . , Y s , and f i are said node data vectors f 1 , . . . , f r for corresponding ones of said word models.

Claim 30

A prefiltering method for processing continuous speech, said speech including a succession of utterances, comprising the steps of: A. storing a plurality of word models; B. identifying a succession of temporal segments of said utterances; C. operating on ones of said segments and selectively identifying a subset of said plurality of word models meeting predetermined criteria, said subset defining a list of candidate words; D. determining arbitrarily selected frame start times t during said utterances, said frame start times being independent of identification of an initial anchor; and E. generating a signal representative of said list of candidate words for selected ones of said frame start times t determined in step D.

Claim 31

A prefiltering method for processing continuous speech, said speech including a succession of utterances, comprising the steps of: A. storing a plurality of word models; B. identifying a succession of temporal segments of said utterances; C. operating on ones of said segments and selectively identifying a subset of said plurality of word models meeting predetermined criteria, said subset defining a list of candidate words; D. determining arbitrarily selected frame start times t during said utterances, said frame start times being independent of identification of an initial anchor; and E. generating a signal representative of said list of candidate words for selected ones of said frame start times t determined in step D.