Patent EP0504927A2 successfully refreshed from Google Patents.
EP0504927A2
Expired / Lapsed
EP
⏱️ 0.0 years remaining
Speech recognition system and method
Assignee: Toshiba Corp
•
Filing: 1992-03-20
•
Granted: —
•
Expires: 2012-03-20
Abstract
Abstract A sound analyzer (11) sound analyzes an input speech signal to obtain feature vectors. A matrix quantizer (13) performs a matrix quantization process between the feature vectors obtained by the sound analyzer and a phonetic segment dictionary prepared in phonetic segment units to obtain a phonetic segment similarity sequence. A phonetic segment integrating section (14) integrates the phonetic segment similarity sequence into a phonemic feature vector. A HMM recognizer (15) checks the phonemic feature vector using a HMM prepared in certain units, to thereby perform a recognition process.
Inventors
- 👤 Tsuneo c/o Intellectual Property Div. Nitta
Bibliographic Details
Application Number
—
Jurisdiction
EP
Data Source
google_patents
Last Synchronized
2026-10-30 13:17
CPC Classifications
G; G09; G09B; G09B19/00; G09B19/04; G10; G10L; G10L15/00; G10L15/08; G10L15/14; G10L15/142; G10L15/144; G10L15/02; G10L2015/025; G10L25/00; G10L25/03; G10L25/12
Claims Analysis
Total Claims: 12
Claim 1
Independent Claim
116 words
A speech recognition system characterized by comprising: sound analyzing means (12) for sound analyzing an input speech signal to obtain feature parameters; a phonetic segment dictionary (13b) for storing a plurality of types of phonetic segments in advance; means (13a) for obtaining a phonetic segment similarity sequence from said feature parameters obtained by said sound analyzing means and said phonetic segment dictionary; integrating means (14) for integrating said phonetic segment similarity sequence obtained by said phonetic segment similarity sequence obtaining means into a phonemic feature vector; and means (15, 17) for checking said phonemic feature vector obtained by said integrating means using a hidden markov model (HMM) prepared in certain units, thereby performing a recognition process.
Claim 2
Dependent Claim
42 words
A system according to claim 1, characterized in that said means for obtaining a phonetic segment similarity sequence comprises matrix quantization means for performing matrix quantization processing between said feature parameters obtained by said sound analyzing means and said phonetic segment dictionary.
Claim 3
Dependent Claim
42 words
A system according to claim 1, characterized in that said means for obtaining a phonetic segment similarity sequence comprises vector quantization means for performing vector quantization processing between said feature parameters obtained by said sound analyzing means and said phonetic segment dictionary.
Claim 4
Dependent Claim
39 words
A system according to claim 1, characterized in that said means for obtaining a phonetic segment similarity sequence comprises a neural network for obtaining the phonetic segment similarity sequence from said feature parameters obtained by said sound analyzing means.
Claim 5
Dependent Claim
27 words
A system according to claim 1, characterized in that said integrating means comprises maximum-value filters for integrating said phonetic segment similarity sequence into a phonemic feature vector.
Claim 6
Dependent Claim
27 words
A system according to claim 1, characterized in that said integrating means comprises neural networks for integrating said phonetic segment similarity sequence into a phonemic feature vector.
Claim 7
Dependent Claim
97 words
A speech recognition method characterized by comprising the steps of: a) analyzing an input speech signal to obtain feature parameters; b) obtaining a phonetic segment similarity sequence from said feature parameters obtained by said step a) and a phonetic segment dictionary (13b) for storing a plurality of types of phonetic segments in advance; c) integrating said phonetic segment similarity sequence obtained by said step b) into a phonemic feature vector; and d) checking said phonemic feature vector obtained by said step c) using a hidden markov model (HMM) prepared in certain units, thereby performing a recognition process.
Claim 8
Dependent Claim
38 words
A method according to claim 7, characterized in that said step b) comprises a matrix quantization step for performing matrix quantization processing between said feature parameters obtained by said sound analyzing step a) and said phonetic segment dictionary.
Claim 9
Dependent Claim
38 words
A method according to claim 7, characterized in that said step b) comprises a vector quantization step for performing vector quantization processing between said feature parameters obtained by said sound analyzing step a) and said phonetic segment dictionary.
Claim 10
Dependent Claim
34 words
A method according to claim 7, characterized in that said step b) uses a neural network to obtain the phonetic segment similarity sequence from said feature parameters obtained by said sound analyzing step a).
Claim 11
Dependent Claim
27 words
A method according to claim 7, characterized in that said step c) uses maximum-value filters to integrate said phonetic segment similarity sequence into a phonemic feature vector.
Claim 12
Dependent Claim
27 words
A method according to claim 7, characterized in that said step c) uses neural networks to integrate said phonetic segment similarity sequence into a phonemic feature vector.
References Cited (Backward Citations)
No citation records indexed.
USPTO Annuity & Maintenance Windows
Maintenance schedule is calculated for granted US utility patents.