Automatic speech/speaker recognition over digital wireless channels

• US 6,336,090 B1
• Filed: 11/30/1998
• Issued: 01/01/2002
• Est. Priority Date: 11/30/1998
• Status: Expired due to Term

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Lucent Technologies, Inc. (Nokia Corporation)
• Inventors: Recchione, Michael Charles, Chou, Wu, Zhou, Qiru
• Primary Examiner(s): Korzuch, William
• Assistant Examiner(s): Storm, Donald L.
• Application Number:
  US09/201,082
• Time in Patent Office: 1,128 Days
• Field of Search: 704/231, 704/243, 704/244, 704/221, 704/222
• US Class Current: 704/231
• CPC Class Codes: G10L 15/02 Feature extraction for speech recognition; Selection of recognition unit
• CPC Class Codes: G10L 15/20 Speech recognition techniques specially adapted for robustness in adverse environments, e.g. in noise, of stress induced speech G10L21/02 takes precedence
• CPC Class Codes: G10L 15/30 Distributed recognition, e.g. in client-server systems, for mobile phones or network applications

Abstract

Automatic Speech Recognition (ASR) is achieved in wireless communications systems in which reliable ASR feature vector sequences are derived at a base station directly from digitally transmitted speech coder parameters, with no additional processing or signal modification required at the originating handset. No secondary channel need be provided for the transmission of ASR feature vectors. In operating on received speech coder parameters prior to conversion to a voice signal the present system and methods avoid the lossy conversion process and associated voice distortion. Since the received voice parameters are error protected during transmission they are received with greater accuracy.

All, or a subset, of speech coding parameters, including, in appropriate cases, spectral envelope parameters, reflection coefficients, LSPs, LSFs, LPCs, LPCCs, and weighted LPCCs may be processed at a receiving base station or forwarded to another location for processing.

Claims

1. A method for automatically recognizing speech received as coded speech signals in a first set of signals received from a wireless transmission channel comprising the steps of

extracting a plurality of speech feature signals from said received coded speech signals separately from any conversion of said received coded speech signals to audio speech signals, and

applying said plurality of speech feature signals to a speech recognition system,

wherein said received coded speech signals comprise a plurality of coded speech parameters and said extracting comprises

selecting a subset of said received coded speech parameters, and

converting said subset of said received coded speech parameters to said plurality of speech feature signals.

2. The method of claim 1, wherein said first set of received signals further comprises error protection signals for improving the reliability of said received coded speech signals.

3. The method of claim 1, wherein said received coded speech parameters comprise a plurality of coded speech spectral envelope parameters for said speech, and wherein said selecting of a subset of said received coded speech parameters comprises selecting a subset of said coded speech spectral envelope parameters.

4. The method of claim 3 wherein said subset of coded speech spectral envelope parameters comprises a plurality of reflection coefficients.

5. The method of claim 3 wherein said subset of coded speech spectral envelope parameters comprises a plurality of line spectral pairs (LSPs).

6. The method of claim 3 wherein said subset of coded speech spectral envelope parameters comprises a plurality of line spectral frequencies (LSFs).

7. The method of claim 3 wherein said converting further compises the step of generating from said subset of coded speech spectral envelope parameters a set of linear predictive coding (LPC) parameters.

8. The method of claim 7 wherein said converting further compises the step of transforming said set of LPC parameters to a corresponding set of ILPC cepstral coefficients.

9. The method of claim 8 wherein said converting further compises the steps of

weighting said set of LPC cepstral coefficients to produce a set of weighted LPC cepstral coefficients, and

deriving said plurality of speech feature signals from said weighted LPC cepstral coefficients.

10. The method of claim 1 wherein said extraction and applying are performed at a base station.

11. The method of claim 1 wherein said first set of signals is received at a base station and forwarded to a second location for said extraction and applying.

12. A system for automatically recognizing speech received as coded speech signals in a first set of signals received from a wireless transmission channel comprising

means for extracting a plurality of speech feature signals from said received coded speech signals separately from any conversion of said received coded speech signals to audio speech signals, and

means for applying said plurality of speech feature signals to a speech recognition system,

wherein said received coded speech signals comprise a plurality of coded speech parameters, and

said means for extracting comprises

means for selecting a subset of said received coded speech parameters, and

means for converting said subset of said received coded speech parameters to said plurality of speech feature signals.

13. The system of claim 12 wherein said received coded speech parameters comprise a plurality of coded speech spectral envelope parameters for said speech, and wherein said means for selecting a subset of said received coded speech parameters comprises means for selecting a subset of said coded speech spectral envelope parameters.

14. The system of claim 13 wherein said subset of said coded speech spectral envelope parameters are parameters of relatively higher reliability by reason of having been protected to a relatively higher degree by error protection means prior to said receiving.

15. The system of claim 12 wherein said first set of signals is received at a base station and said means for extracting and means for applying are located at said base station.

16. The system of claim 12 wherein said first set of signals is received at a base station and said system further comprises means for forwarding said first set of signals to a second location for said extraction and applying.