Distributed voice recognition system

• US 6,594,628 B1
• Filed: 04/02/1997
• Issued: 07/15/2003
• Est. Priority Date: 09/21/1995
• Status: Expired due to Term

• Current Assignee: Qualcomm, Inc.
• Original Assignee: Qualcomm, Inc.
• Inventors: Jacobs, Paul E., Chang, Chienchung
• Primary Examiner(s): To, Doris H.
• Assistant Examiner(s): Opsasnick, Michael N.
• Application Number:
  US08/832,581
• Time in Patent Office: 2,295 Days
• Field of Search: 704/231, 704/246, 704/270, 704/275, 704/201, 455/461, 455/563, 379/88.01, 379/88.02
• US Class Current: 704/231
• CPC Class Codes: G10L 15/02 Feature extraction for speech recognition; Selection of recognition unit
• CPC Class Codes: G10L 15/30 Distributed recognition, e.g. in client-server systems, for mobile phones or network applications
• CPC Class Codes: G10L 19/00 Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis in musical instruments G10H

Abstract

A voice recognition system is disclosed wherein a feature extraction apparatus is located in a remote station. The feature extraction apparatus extracts features from an input speech frame and then provides the extracted features to a central processing station. In the central processing station, the features are provided to a word decoder which determines the syntax of the input speech frame.

Claims

1. A method for providing voice recognition in a wireless communication system, the method comprising the steps of:

providing a speech signal spoken by a user to a feature extractor located at a subscriber station;

extracting a set of speech features from the speech signal in accordance with a predetermined feature extraction format;

determining a speech meaning from the extracted speech features in accordance with a small vocabulary that is locally stored at the subscriber station; and

if the speech meaning cannot be decoded by a first word decoder located at the at the subscriber station in accordance with the small vocabulary, then transmitting the extracted speech features to a central processing station using a second word decoder located at the central processing station to decode the extracted speech features in accordance with a larger vocabulary.

2. The method recited in claim 1 wherein the set of speech features comprises linear spectrum pair (LSP) parameter data.

3. The method recited in claim 2 further comprises converting the LSP parameter data into linear predictive coding (LPC) parameter data.

4. The method recited in claim 1, wherein the set of speech features are formed by encoding the speech signal in accordance with a variable rate code excited linear prediction (CELP) encoding format.

5. A method for providing a distributed voice recognition system, the method comprising the steps of:

processing a frame of speech samples at a subscriber station to determine acoustic features of the frame of speech samples;

using a first word decoder located at the subscriber station to decode the acoustic features, wherein the first word decoder uses a small vocabulary to decode the acoustic features;

transmitting the acoustic features from the subscriber station to a central processing station if the acoustic features cannot be decoded by the first word decoder;

receiving the acoustic features at the central processing station;

using a second word decoder located at the central processing station to decode the acoustic features, wherein the remote word decoder uses a larger vocabulary to decode the acoustic features; and

transmitting the decoded acoustic features from the central processing station to the subscriber station.

6. The method recited in claim 5 wherein processing the frame of speed samples to determine acoustic features comprises extracting the acoustic features of the frame of speech samples in accordance with a predetermined feature extraction format.

7. A method for providing a distributed voice recognition system, the method comprising the steps of:

extracting acoustic features of a frame of speech samples at a subscriber station;

decoding the acoustic features in accordance with a small vocabulary using a first word decoder located at the subscriber station;

transmitting the extracted acoustic features from the subscriber station to a central processing station if the extracted acoustic features cannot be decoded by the first word decoder;

receiving the extracted acoustic features at the central processing station; and

decoding the extracted acoustic features in accordance with a larger vocabulary using a second word decoder located at the central processing station,

if the decoded acoustic features are associated with a local command signal, then transmitting the local command signal from the central processing station to the subscriber station.