§
Luxuse
Patent Intelligence
Sign In
Patent EP0316112A2 successfully refreshed from Google Patents.
← Back to Patent Explorer
EP0316112A2 Expired / Lapsed EP ⏱️ 0.0 years remaining

Use of instantaneous and transitional spectral information in speech recognizers

Assignee: AT&T Inc
•
Filing: 1988-11-03
•
Granted: —
•
Expires: 2008-11-03
📄 Original PDF 🌐 Google Patents 🔄 Sync Now

Abstract

Abstract Any of several prior art word recognizers are improved by changing the speech analysis portion thereof to include a more comprehensive spectral representation which can be characterized as two-dimensional. The first dimension is the "freeze-frame" or instantaneous sample of the spectral characteristics as derived from a typical 45 millisecond sample of the speech to be recognized. The second dimension, in a sense orthogonal to the first, spans several such time frames or samples and yields what amounts to a time derivative of the spectral properties obtained in the current time frame. In the testing or processing of the input speech in comparison to the reference patterns, essentially equal weight or effectiveness is attributed to both parts of the two dimensional reference spectral pattern and to both parts of the input speech spectral pattern. That is, equal importance is attributed to the essentially instantaneous information and the time-derivative information spanning a plurality of neighboring time frames including the frame of interest. Also disclosed are the use of cepstral information and time-derivative cepstral information derived from linear prediction coefficient methods and the use specifically in connected-word recognizers using level-building concepts, or in those using two-stage processing.

Inventors

  • 👤 Lawrence Richard Rabiner
  • 👤 Frank Kao-Ping Soong
  • 👤 Jay Gordon Wilpon

Bibliographic Details

Application Number —
Jurisdiction EP
Data Source google_patents
Last Synchronized 2026-11-02 02:28
CPC Classifications
G; G10; G10L; G10L15/00

Claims Analysis

Total Claims: 5
Claim 1
Independent Claim 185 words
A speech recognition system of the type comprising means for storing reference patterns derived from speech in a form useful for comparison with a like form derived from speech to be recognized, means for deriving from the speech to be recognized patterns in the like form with regard to the reference patterns, and means for processing the respective patterns to generate probable matches of the patterns, and means for indicating recognized words for utilization, said system BEING CHARACTERIZED IN THAT the storing means includes means for storing in an associated form sets of patterns having both data from a single time frame of the detected speech and transitional data spanning a plurality of neighboring time frames including the current frame, the deriving means comprises means for deriving, in an associated form, sets of patterns having both data from a single time frame of the detected speech, and transitional data spanning a plurality of neighboring time frames including the single time frame, and the processing means includes means for giving equivalent weight to both single time frame data and transitional data in obtaining probable recognized words.
Claim 2
Dependent Claim 185 words
A speech recognition system of the type comprising means for storing reference patterns derived from speech in a form useful for comparison with a like form derived from speech to be recognized, means for deriving from the speech to be recognized patterns in the like form with regard to the reference patterns, and means for processing the respective patterns to generate probable matches of the patterns, and means for indicating recognized words for utilization, said system BEING CHARACTERIZED IN THAT the storing means includes means for storing in an associated form sets of patterns having both data from a single time frame of the detected speech and transitional data spanning a plurality of neighboring time frames including the current frame, the deriving means comprises means for deriving, in an associated form, sets of patterns having both data from a single time frame of the detected speech, and transitional data spanning a plurality of neighboring time frames including the single time frame, and the processing means includes means for giving equivalent weight to both single time frame data and transitional data in obtaining probable recognized words.
Claim 3
Dependent Claim 66 words
A speech recognition system of the type claimed in claim 1, said system BEING FURTHER CHARACTERIZED IN THAT the storing means includes means for storing, and the deriving means includes means for deriving, sets of patterns including as the transitional data respective time-derivative data centered on the respective single frames, wherein the time-derivative data depends in part on data from one or more neighboring time frames.
Claim 4
Dependent Claim 136 words
A speech recognition system of the type claimed in claim 1, said system BEING FURTHER CHARACTERIZED IN THAT the means for storing the reference patterns comprises means for deriving spectral features by Linear Prediction Coefficient methods the single-­time-frame data comprising cepstral information and the transitional data comprising time-derivative cepstral information characterizing the neighborhood of related single-time-frame data, the means for deriving patterns from the speech to be recognized comprises means for deriving spectral features by Linear Prediction Coefficient (LPC) methods, the single-time-frame data comprising cepstral information, and the transitional data comprising time-derivative cepstral information characterizing the neighborhood of the related single-time-frame data, and the means for processing the respective patterns includes means for applying the single-time-frame cepstral information and the time-derivative cepstral information equally effectively, to improve the likelihood of accuracy in recognizing words for utilization.
Claim 5
Dependent Claim 89 words
A speech recognition system of the type claimed in claim 3, said system BEING FURTHER CHARACTERIZED IN THAT the means for storing reference patterns comprises means for storing such patterns derived by a segmental training procedure from samples of connected-word speech, the means for deriving patterns from the speech to be recognized is adapted to act upon connected-word speech, and and the means for processing the respective patterns is adapted to apply both the cepstral information and the time-derivative cepstral information effectively in each level of a level-building process.

References Cited (Backward Citations)

No citation records indexed.

USPTO Annuity & Maintenance Windows

Maintenance schedule is calculated for granted US utility patents.

Assigned Categories & Tags

No categories assigned to this patent yet.

Assign Category Tag