US4723290A
Expired / Lapsed
US
⏱️ 0.0 years remaining
Speech recognition apparatus
Assignee: Toshiba Corp
•
Filing: 1984-05-08
•
Granted: 1988-02-02
•
Expires: 2004-05-08
Abstract
Abstract Speech recognition is improved by first recognizing a sequence of phonemes, then generating a compatibility coefficient between two continuous phonemes to modify the similarity measurements. The compatibility coefficient is varied for a plurality of modified similarity measurements to further improve recognition accuracy.
Inventors
- 👤 Sadakazu Watanabe
- 👤 Teruhiko Ukita
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-10-25 20:29
CPC Classifications
G; G10; G10L; G10L15/00; G10L15/08; G10L15/10; G10L15/02; G10L2015/025
Claims Analysis
Total Claims: 7
Claim 1
Independent Claim
187 words
A speech recognition apparatus comprising: parameter generating means for sequentially generating acoustic parameter data corresponding to input speech; data generating means for detecting a segment in which each discrete phonetic symbol exists on the basis of acoustic parameter data from said parameter generating means and generating phonetic symbol data corresponding to at least one reference phonetic symbol having a similarity larger than a predetermined value with respect to the discrete phonetic symbol in each segment and score data corresponding to the similarity; memory means for storing the phonetic symbol data and score data from said data generating means; coefficient generating means connected to said parameter generating means and data generating means and for generating compatability coefficient between phonetic symbols generated in two continuous segments; updating means for updating plural times the score data in said memory means with compatibility coefficient relating to the corresponding phonetic symbol data as a function; and recognition means for sequentially reading out from said memory means phonetic symbol data having the maximum score data in each segment after the updating process is finished, and then recognizing the sequence of phonetic symbol data.
Claim 2
Dependent Claim
187 words
A speech recognition apparatus comprising: parameter generating means for sequentially generating acoustic parameter data corresponding to input speech; data generating means for detecting a segment in which each discrete phonetic symbol exists on the basis of acoustic parameter data from said parameter generating means and generating phonetic symbol data corresponding to at least one reference phonetic symbol having a similarity larger than a predetermined value with respect to the discrete phonetic symbol in each segment and score data corresponding to the similarity; memory means for storing the phonetic symbol data and score data from said data generating means; coefficient generating means connected to said parameter generating means and data generating means and for generating compatability coefficient between phonetic symbols generated in two continuous segments; updating means for updating plural times the score data in said memory means with compatibility coefficient relating to the corresponding phonetic symbol data as a function; and recognition means for sequentially reading out from said memory means phonetic symbol data having the maximum score data in each segment after the updating process is finished, and then recognizing the sequence of phonetic symbol data.
Claim 3
Dependent Claim
38 words
A speech recognition apparatus according to claim 1, wherein said compatibility coefficient generating means comprises a memory for storing a plurality of relative recognition coefficient data respectively representing the compatibility coefficients between two preselected continuous phoneme symbol data.
Claim 4
Dependent Claim
144 words
A speech recognition apparatus according to claim 2, wherein said data generating means comprises a segment detecting circuit connected to said parameter generating means and for detecting each segment on the basis of the acoustic parameter data from said parameter generating means, reference pattern generating means for generating a plurality of reference pattern data representing a plurality of reference phonetic symbols, and score calculating circuit connected to said segment detecting circuit and reference pattern generating means and for generating corresponding phonetic symbol data and score data when the score having a value higher than a predetermined value is obtained by calculating the score corresponding to the similarity between the acoustic pattern data composed of the acoustic parameter data generated from said parameter generating means and the reference pattern data from said reference pattern generating means in each segment detected by said segment detecting circuit.
Claim 5
Dependent Claim
111 words
A speech recognition apparatus according to claim 3, wherein said coefficient generating means comprises a memory for storing a plurality of reference boundary pattern data respectively representing acoustic features between two continuous phonetic symbols, score data generator connected to said memory and parameter generating means and for generating score data corresponding to the similarity between boundary pattern data composed of a predetermined number of acoustic parameter data generated from said parameter generating means between two successive segments and reference boundary pattern data read out from said memory in accordance with the phonetic symbol data in these continuous segments, and a second memory for storing score data from said score data generator.
Claim 6
Dependent Claim
125 words
A speech recognition apparatus according to claim 4, wherein said data generating means comprises a segment detecting circuit for detecting each segment on the basis of the acoustic parameter data from said parameter generating means, reference pattern generating means for generating a plurality of reference pattern data representing a plurality of reference phonetic symbols, and score calculating circuit for generating corresponding phonetic symbol data and score data when the score having a value higher than a predetermined value is obtained by calculating the score corresponding to the similarity between the acoustic pattern data composed of the acoustic parameter data generated from said parameter generating means and the reference pattern data from said reference pattern generating means in each segment detected by said segment detecting circuit.
Claim 7
Dependent Claim
125 words
A speech recognition apparatus according to claim 1, wherein said data generating means comprises a segment detecting circuit for detecting each segment on the basis of the acoustic parameter data from said parameter generating means, reference pattern generating means for generating a plurality of reference pattern data representing a plurality of reference phonetic symbols, and score calculating circuit for generating corresponding phonetic symbol data and score data when the score having a value higher than a predetermined value is obtained by calculating the score corresponding to the similarity between the acoustic pattern data composed of the acoustic parameter data generated from said parameter generating means and the reference pattern data from said reference pattern generating means in each segment detected by said segment detecting circuit.
References Cited (Backward Citations)
USPTO Annuity & Maintenance Windows
| Annuity Stage | Window Opens | Payment Deadline | Surcharge Period Until | Estimated Fee | Status |
|---|---|---|---|---|---|
| 3.5 Years (Year 4) | 1991-02-01 | 1991-08-03 | 1992-02-02 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 1995-02-01 | 1995-08-03 | 1996-02-02 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 1999-02-01 | 1999-08-03 | 2000-02-02 | $7700.00 | Window Closed / Lapsed |