EP0181167A2
Expired / Lapsed
EP
⏱️ 0.0 years remaining
Apparatus and method for identifying spoken words
Assignee: Alcatel Lucent NV
•
Filing: 1985-11-01
•
Granted: —
•
Expires: 2005-11-01
Abstract
Abstract An apparatus and method for identifying spoken words employs a plurality of reference templates (26) having information at each position thereof representing the probability of a particular binary value occurring at a corresponding position in an utterance template (16). In addition, a system bias template (42) is included to substantially eliminate recognition inaccuracies characteristic to the system.
Inventors
- 👤 Priyadarshan Jakatdar
- 👤 Hoshang Dadi Mulla
Bibliographic Details
Application Number
—
Jurisdiction
EP
Data Source
google_patents
Last Synchronized
2026-10-27 20:26
CPC Classifications
G; G10; G10L; G10L15/00; G10L15/06; G10L15/063
Claims Analysis
Total Claims: 31
Claim 1
Independent Claim
141 words
An apparatus for speech recognition, characterised in an utterance template (16) having a plurality of positions, each position having a binary value stored therein, each binary value in each position representing a spectral feature of speech to be recognised, a plurality of reference templates (26), each having a plurality of positions, the positions of each reference template corresponding to the plurality of positions of the utterance template, each position in each reference template having a value stored therein, which stored value is representative of the probability of a particular binary value occurring in the corresponding position of the utterance template, and means (22) for establishing a first score for each reference template, each of which scores is indicative of the relative match between that reference template and the utterance template whereby the speech represented by the utterance template can be recognised.
Claim 2
Independent Claim
141 words
An apparatus for speech recognition, characterised in an utterance template (16) having a plurality of positions, each position having a binary value stored therein, each binary value in each position representing a spectral feature of speech to be recognised, a plurality of reference templates (26), each having a plurality of positions, the positions of each reference template corresponding to the plurality of positions of the utterance template, each position in each reference template having a value stored therein, which stored value is representative of the probability of a particular binary value occurring in the corresponding position of the utterance template, and means (22) for establishing a first score for each reference template, each of which scores is indicative of the relative match between that reference template and the utterance template whereby the speech represented by the utterance template can be recognised.
Claim 3
Dependent Claim
50 words
An apparatus as claimed in claim 1, characterised in that the first score establishing means (22) includes means (88A, 88B) for summing outputs from the reference template (26), each output corresponding to one of the plurality of positions of the reference template and being determined from the stored value therein.
Claim 4
Dependent Claim
88 words
An apparatus as claimed in claim 1 or 2, characterised in a bias template (42) having a plurality of positions corresponding to the plurality of positions of the utterance template (16), each position of the bias template having a value stored therein, the stored value being representative of the probability of a particular binary value occurring in the corresponding position of any utterance template, and means for establishing a second score, the second score being indicative of the relative match between the bias template and the utterance template.
Claim 5
Dependent Claim
48 words
An apparatus as claimed in claim 3, characterised in that the second score establishing means includes means (66) for summing outputs from the bias template (42), each output corresponding to one of the plurality of position of the bias template and being determined from the stored value therein.
Claim 6
Dependent Claim
152 words
An apparatus as claimed in claim 3 or 4, characterised in means for expanding each reference template (26) into a first expanded reference template (82) and a second expanded reference template (84), the values of the first expanded reference template representing the probability of a binary one occurring in the corresponding position of the utterance template and the values of the second expanded reference template representing the probability of a binary zero occurring at the corresponding position of the utterance template, and means for expanding the bias template (42) into a first expanded bias template (70) and a second expanded bias template (72) the values of the first expanded bias template representing the probability of a binary one occurring in the corresponding position of any utterance template and the values of the second expanded bias template representing the probability of a binary zero occurring at the corresponding position of any utterance template.
Claim 7
Dependent Claim
44 words
An apparatus as claimed in claim 5, characterised in that the first score establishing means includes means (102) for selecting either the value in the first expanded reference template (82) or the value in the second expanded reference template (84) for scoring each position.
Claim 8
Dependent Claim
46 words
An apparatus as claimed in claim 5 or 6, characterised in that the second score establishing means includes means (100) for selecting either the value in the first expanded bias template (70) or the value in the second expanded bias template (72) for scoring each position.
Claim 9
Independent Claim
53 words
An apparatus as claimed in any one of claims 5 to 7, characterised in that the expanding means includes means (74, 62) for logarithmically converting the values in the plurality of reference templates (26) and the values in the bias template (42) whereby the first and second score establishing means perform only addition.
Claim 10
Independent Claim
55 words
An apparatus as claimed in any one of claims 5 to 8, characterised in means for comparing the highest of the first score with a threshold, and means (116) for generating a recognition acceptance signal when the highest score exceeds the threshold and a rejection signal when the highest score is less than the threshold.
Claim 11
Dependent Claim
45 words
An apparatus as claimed in claim 3, characterised in means for comparing the second score with a second threshold, and means (116) for generating a recognition acceptance signal only when the highest score exceeds the threshold and when the second score exceeds the second threshold.
Claim 12
Dependent Claim
45 words
An apparatus as claimed in claim 1, characterised in means for enrolling the plurality of reference templates, each reference template being enrolled from a plurality of utterance templates; each plurality of utterance templates being composed of spectral features representing different instances of the same speech.
Claim 13
Dependent Claim
30 words
An apparatus as claimed in claim 11, characterised in means for enrolling a bias template, the bias template being enrolled from all utterance templates employed to enroll the reference templates.
Claim 14
Independent Claim
141 words
an apparatus for use in speech recognition, characterised in an utterance template (16), having a plurality of positions, each position having a binary value stored therein, each binary value representing a spectral feature of speech to be recognised, a plurality of reference templates (26), each template having positions corresponding to the plurality of positions, each position having a reference value stored therein, a bias template (42) having a plurality of positions, each one of the plurality of positions corresponding to one of the plurality of position in the utterance template, each position of the bias template having a value stored therein representing the probability of a particular binary value occurring at the corresponding position in any utterance template, and means (22) for establishing a score for each reference template based on the stored reference values therein and said bias template values.
Claim 15
Independent Claim
141 words
an apparatus for use in speech recognition, characterised in an utterance template (16), having a plurality of positions, each position having a binary value stored therein, each binary value representing a spectral feature of speech to be recognised, a plurality of reference templates (26), each template having positions corresponding to the plurality of positions, each position having a reference value stored therein, a bias template (42) having a plurality of positions, each one of the plurality of positions corresponding to one of the plurality of position in the utterance template, each position of the bias template having a value stored therein representing the probability of a particular binary value occurring at the corresponding position in any utterance template, and means (22) for establishing a score for each reference template based on the stored reference values therein and said bias template values.
Claim 16
Dependent Claim
37 words
An apparatus as claimed in claim 13, characterised in that each reference value stored in each reference template (26) represent the probability of a particular binary value occurring in a corresponding position in the utterance template (16).
Claim 17
Dependent Claim
103 words
An apparatus as claimed in claim 13 or 14, characterised in means for expanding the bias template (42) into a first and a second bias template (70, 72), each having positions corresponding to the plurality of positions in the utterance templates (16), the first bias template (70) having bias values stored in each position therein representing the probability of a binary one occurring at the corresponding position in the utterance template (16) and the second bias template (72) having bias values stored in each position therein representing the probability of a binary zero occurring at the corresponding position in the utterance template (16).
Claim 18
Independent Claim
90 words
a method of matching templates, characterised in the steps of providing an utterance template having a plurality of positions each having a binary value stored therein, providing a plurality of reference templates each having a plurality of positions corresponding to the plurality of positions in the utterance template and having a value stored therein, each stored value representing the probability of a particular binary value occurring in the corresponding position in the utterance template, and establishing a first score for each reference template whereby the utterance template can be matched.
Claim 19
Independent Claim
90 words
a method of matching templates, characterised in the steps of providing an utterance template having a plurality of positions each having a binary value stored therein, providing a plurality of reference templates each having a plurality of positions corresponding to the plurality of positions in the utterance template and having a value stored therein, each stored value representing the probability of a particular binary value occurring in the corresponding position in the utterance template, and establishing a first score for each reference template whereby the utterance template can be matched.
Claim 20
Dependent Claim
26 words
A method as claimed in claim 16, characterised in that the first score establishing step includes summing the outputs from each position of each reference template.
Claim 21
Dependent Claim
78 words
A method as claimed in claim 16 or 17, characterised in the steps of providing a bias template having a plurality of positions corresponding to the plurality of positions of the utterance template each position having a value stored therein, the stored value representing the probability of a particular binary value occurring in the corresponding position in any utterance template, and establishing a second score indicative of the relative match between the bias template and the utterance template.
Claim 22
Dependent Claim
26 words
A method as claimed in claim 18, characterised in that the second score establishing step includes summing outputs from the bias template for each position thereof.
Claim 23
Dependent Claim
143 words
A method as claimed in claim 19, characterised in the steps of expanding each reference template into a first expanded reference template and a second expanded reference template, the values of the first expanded reference template representing the probability of a binary one occurring in the corresponding position of the utterance template and the values of the second expanded reference template representing the probability of a binary zero occurring at the corresponding position of the utterance template, and expanding the bias template into a first expanded bias template and a second expanded bias template, the values of the first expanded bias template representing the probability of a binary one occurring in the corresponding position of any utterance template and the values of the second expanded bias template representing the probability of a binary zero occurring at the corresponding position of any utterance template.
Claim 24
Dependent Claim
39 words
A method as claimed in claim 20, characterised in that the first score establishing step includes selecting either the value in the first expanded reference template or the value in the second expanded reference template for scoring each position.
Claim 25
Dependent Claim
39 words
A method as claimed in claim 21, characterised in that the second score establishing step includes selecting either the value in the first expanded bias template or the value in the second expanded bias template for scoring each position.
Claim 26
Independent Claim
46 words
A method as claimed in any one of claims 20 to 22, characterised in the step of converting the values in the plurality of reference templates and the values in the bias template to data whereby the first and second score establishing means perform only addition.
Claim 27
Dependent Claim
32 words
A method as claimed in claim 20, characterised in that the conversion of the values in the plurality of reference templates and the values in the bias template are carried out logarithmically.
Claim 28
Dependent Claim
50 words
A method as claimed in claim 23 or 24, characterised in the steps of comparing the highest of the first scores with a threshold, and generating a recognition acceptance signal when the highest score exceeds the threshold and a rejection signal when the highest score is less than the threshold.
Claim 29
Independent Claim
48 words
A method as claimed in any one of claims 23 to 25, characterised in the steps of comparing the second score with a second threshold, and generating a recognition acceptance signal only when the highest score exceeds the threshold and when the second score exceeds the second threshold.
Claim 30
Dependent Claim
49 words
A method as claimed in claim 16, characterised in the step of enrolling the plurality of reference templates, each reference template being enrolled from a plurality of utterance templates, each of the plurality of utterance templates being composed of spectral features representing the different instances of the same speech.
Claim 31
Dependent Claim
33 words
A method as claimed in claim 27, characterised in the step of enrolling a bias template, the bias template being enrolled from all of the utterance templates employed to enroll the reference templates.
References Cited (Backward Citations)
No citation records indexed.
USPTO Annuity & Maintenance Windows
Maintenance schedule is calculated for granted US utility patents.