US4991216A
Expired / Lapsed
US
⏱️ 0.0 years remaining
Method for speech recognition
Assignee: Panasonic Holdings Corp
•
Filing: 1990-03-23
•
Granted: 1991-02-05
•
Expires: 2010-03-23
Abstract
Abstract Phoneme standard patterns are prepared beforehand using speech sounds from a number of individual speakers. Unknown input speech sounds are divided into continuous frames and then some of these frames are extracted such that extracted frames include noncontinuous frames. LPC cepstrum coefficients are obtained for each of the frames and a mean value is obtained from the LPC cepstrum coefficients for each of the phonemes. A first standard pattern is then formed for each of the phonemes belonging to each group determined by characteristics of known speech sounds. Spectrum information of the unknown input speech sounds is produced using such extracted frames, and is then compared with the phoneme standard patterns prestored in a storage to determine and recognize phonemes of the unknown speech sounds by calculating similarity between the two using a statistical distance measure. When obtaining LPC cepstrum coefficients through LPC analysis, the order of LPC cepstrum coefficients is set to a value below the order of LPC analysis. The method according to the present invention reduces the amount of calculations necessary to perform speech recognition without deteriorating phoneme and word recognition rate.
Inventors
- 👤 Satoru Fujii
- 👤 Katsuyuki Niyada
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-11-08 00:42
CPC Classifications
G; G10; G10L; G10L15/00; G10L15/08; G10L15/10; G10L15/02; G10L2015/025
Claims Analysis
Total Claims: 10
Claim 1
Independent Claim
128 words
A method recognizing speech comprising the steps of: (a) dividing speech sounds into continuous N frames each having a predetermined time period; (b) analyzing speech sounds for each of said N frames so as to obtain spectrum information; (c) extracting n frames from said N frames such that said n frames include frames noncontinuous in said continuous N frames wherein n<N, said noncontinuous n frames being extracted from a region where variation rate of distance determined by said spectrum information and a standard pattern is below a predetermined value; and (d) recognizing phonemes or syllables of said speech sounds by calculating similarity between said spectrum information of said n frames and said standard patterns, which are produced using speech sounds from a number of speaking persons in advance.
Claim 2
Independent Claim
128 words
A method recognizing speech comprising the steps of: (a) dividing speech sounds into continuous N frames each having a predetermined time period; (b) analyzing speech sounds for each of said N frames so as to obtain spectrum information; (c) extracting n frames from said N frames such that said n frames include frames noncontinuous in said continuous N frames wherein n<N, said noncontinuous n frames being extracted from a region where variation rate of distance determined by said spectrum information and a standard pattern is below a predetermined value; and (d) recognizing phonemes or syllables of said speech sounds by calculating similarity between said spectrum information of said n frames and said standard patterns, which are produced using speech sounds from a number of speaking persons in advance.
Claim 3
Dependent Claim
30 words
A method as claimed in claim 1, wherein said spectrum information is obtained through one of linear prediction analysis, a set of band pass filters, or Fast Fourier Transform analysis.
Claim 4
Dependent Claim
18 words
A method as claimed in claim 1, wherein the similarity calculation is effected using a statistical distance measure.
Claim 5
Dependent Claim
27 words
A method as claimed in claim 3, wherein said statistical distance measure is one of a distance based on Bayes' discriminant, Mahalanobis' distance, and linear discriminant function.
Claim 6
Independent Claim
353 words
A method of recognizing speech, comprising the steps of: (a) receiving known speech sounds including given phonemes from a number of unspecified speaking persons; (b) dividing said known speech sounds into a plurality of frames each having a predetermined time period; (c) obtaining LPC cepstrum coefficients, for each of said frames, from said known input speech sounds through linear prediction analysis for each of said given phonemes; (d) obtaining a mean value from said LPC cepstrum coefficients for each of said phonemes; (e) producing a first standard pattern formed of spectrum information of a plurality of said frames using said mean value for each of said phonemes or for each of said phonemes belonging to each group determined by characteristics of said known speech sounds; (f) receiving known speech sounds including vowels of said phonemes from a speaking person; (g) dividing said known speech sounds from said speaking person into a plurality of frames each having a predetermined time period; (h) obtaining second LPC cepstrum coefficients, for each of said frames, from said known input speech sounds of said speaking person through linear prediction analysis; (i) obtaining a new mean value from said second LPC cepstrum coefficients for each of said phonemes; (j) producing a new standard pattern by combining said new mean value with said mean value used for producing said first standard pattern thereby correcting said first standard pattern; (k) subsequent to step (j), receiving unknown speech sounds from said speaking person; (1) dividing said unknown speech sounds from said speaking person into a plurality of frames; (m) obtaining third LPC cepstrum coefficients, for each of said frames, from said unknown input speech sounds of said specified speaking person through linear prediction analysis; (n) obtaining similarity or phoneme strings on the basis of Mahalanobis' distance using said second and third LPC cepstrum coefficients; (o) effecting segmentation of phonemes using said similarity or phoneme strings, thereby producing phoneme strings with determination being made for each period obtained by said segmentation; and (p) recognizing a word by comparing said phoneme string with words in a word dictionary expressed in terms of phoneme strings.
Claim 7
Independent Claim
353 words
A method of recognizing speech, comprising the steps of: (a) receiving known speech sounds including given phonemes from a number of unspecified speaking persons; (b) dividing said known speech sounds into a plurality of frames each having a predetermined time period; (c) obtaining LPC cepstrum coefficients, for each of said frames, from said known input speech sounds through linear prediction analysis for each of said given phonemes; (d) obtaining a mean value from said LPC cepstrum coefficients for each of said phonemes; (e) producing a first standard pattern formed of spectrum information of a plurality of said frames using said mean value for each of said phonemes or for each of said phonemes belonging to each group determined by characteristics of said known speech sounds; (f) receiving known speech sounds including vowels of said phonemes from a speaking person; (g) dividing said known speech sounds from said speaking person into a plurality of frames each having a predetermined time period; (h) obtaining second LPC cepstrum coefficients, for each of said frames, from said known input speech sounds of said speaking person through linear prediction analysis; (i) obtaining a new mean value from said second LPC cepstrum coefficients for each of said phonemes; (j) producing a new standard pattern by combining said new mean value with said mean value used for producing said first standard pattern thereby correcting said first standard pattern; (k) subsequent to step (j), receiving unknown speech sounds from said speaking person; (1) dividing said unknown speech sounds from said speaking person into a plurality of frames; (m) obtaining third LPC cepstrum coefficients, for each of said frames, from said unknown input speech sounds of said specified speaking person through linear prediction analysis; (n) obtaining similarity or phoneme strings on the basis of Mahalanobis' distance using said second and third LPC cepstrum coefficients; (o) effecting segmentation of phonemes using said similarity or phoneme strings, thereby producing phoneme strings with determination being made for each period obtained by said segmentation; and (p) recognizing a word by comparing said phoneme string with words in a word dictionary expressed in terms of phoneme strings.
Claim 8
Dependent Claim
101 words
A method as claimed in claim 5, wherein said step of obtaining similarity of phoneme strings comprises the steps of: (a) dividing said unknown speech sounds into a plurality of continuous frames and obtaining spectrum information for each frame; and (b) extracting a plurality of frames from said continuous frames such that frames noncontinuous in said continuous frames are included, and obtaining spectrum information therefor as said spectrum information of said unknown speech sounds, said noncontinuous frames being extracted from a region where variation rate of distance determined by said spectrum information and said standard pattern is below a predetermined value.
Claim 9
Dependent Claim
38 words
A method as claimed in claim 5, wherein said spectrum information is obtained through linear prediction analysis, the order of said LPC cepstrum coefficients being set to a value lower than the order of said linear prediction analysis.
Claim 10
Dependent Claim
48 words
A method as claimed in claim 5, wherein each of said steps of producing said first standard pattern and producing said new standard pattern comprises a step of obtaining a covariance matrix using said mean value, and a step of obtaining an inverse matrix using said covariance matrix.
References Cited (Backward Citations)
US4092493A
Cited
US4400828A
Cited
US4601054A
Cited
US4624010A
Cited
JPS58145998A
Cited
US4592085A
Cited
US4625287A
Cited
US4618984A
Cited
USPTO Annuity & Maintenance Windows
| Annuity Stage | Window Opens | Payment Deadline | Surcharge Period Until | Estimated Fee | Status |
|---|---|---|---|---|---|
| 3.5 Years (Year 4) | 1994-02-04 | 1994-08-06 | 1995-02-05 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 1998-02-04 | 1998-08-06 | 1999-02-05 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 2002-02-04 | 2002-08-06 | 2003-02-05 | $7700.00 | Window Closed / Lapsed |