Speech recognition using both time encoding and HMM in parallel

• US 6,301,562 B1
• Filed: 04/27/2000
• Issued: 10/09/2001
• Est. Priority Date: 04/27/1999
• Status: Expired due to Fees

• Current Assignee: New Transducers Limited (Alphabet Inc.)
• Original Assignee: New Transducers Limited (Alphabet Inc.)
• Inventors: Kavanagh, Sean, Ferekidis, Charalampos, Azima, Henry
• Primary Examiner(s): Dorvil, Richemond
• Assistant Examiner(s): Nolan, Daniel A.
• Application Number:
  US09/559,760
• Time in Patent Office: 530 Days
• Field of Search: 704/251, 704/257, 704/231, 704/211
• US Class Current: 704/257
• CPC Class Codes: G10L 15/02 Feature extraction for speech recognition; Selection of recognition unit

Abstract

A speech recognition method that combines time encoding and hidden Markov approaches. The speech is input and encoded using time encoding, such as TESPAR. A hidden Markov model generates scores; the scores are used to determine the speech element; and the result is output.

Claims

1. A speech recognition method including the steps of

inputting speech to be recognised,

encoding the input speech using time encoding,

using a hidden Markov model to determine scores indicating how the input speech matches some or all of a plurality of speech elements,

determining which, if any, speech element best corresponds to the input speech using both the time encoded speech and the Markov scores, and

outputting the speech element, if any, so determined.

2. A speech recognition method according to claim 1, wherein the time encoding step includes

identifying the intervals between the occurrences of the input crossing a given value, and quantising the lengths of the intervals,

identifying the number of complex zeroes of the input parameter, up to a predetermined rank, in the said intervals, and

recording the quantised lengths of the intervals and a measure of the said number complex zeroes up to a predetermined rank as a representation of the variation of the input parameter.

3. A method according to claim 2, wherein the predetermined rank is 1.

4. A method according to claim 3, wherein the coding method is a TESPAR method.

5. A method according to claim 4, further comprising the step of generating a code number taking one of a set of predetermined values representing the duration of the interval and number of maxima and minima for at least some of the said intervals.

6. A method according to claim 5, wherein the step of determining the best corresponding speech element uses linguistic analysis of previously output speech elements.

7. A method according to claim 3, further comprising the step of generating a code number taking one of a set of predetermined values representing the duration of the interval and number of maxima and minima for at least some of the said intervals.

8. A method according to claim 7, wherein the step of determining the best corresponding speech element uses linguistic analysis of previously output speech elements.

9. A method according to claim 2, further comprising the step of generating a code number taking one of a set of predetermined values representing the duration of the interval and number of maxima and minima for at least some of the said intervals.

10. A method according to claim 9, wherein the step of determining the best corresponding speech element uses linguistic analysis of previously output speech elements.

11. A method according to claim 2, wherein the step of determining the best corresponding speech element uses linguistic analysis of previously output speech elements.

12. A method according to claim 1, wherein the step of determining the best corresponding speech element uses linguistic analysis of previously output speech elements.

13. A speech recognition system comprising

speech capture system for inputting speech to be recognised,

a hidden Markov speech recognition system for determining scores indicating how the input speech matches some or all of a plurality of speech elements,

a time encoded speech system for encoding the input speech, and

a decision system for determining which, if any, speech element best corresponds to the input speech using both the time encoded speech and the Markov scores.

14. A computer program recorded on a data carrier operable to control a computer system having a processing unit, a speech input and an output system, the program operable to control the computer system to carry out the method of

inputting speech to be recognised,

encoding the input speech using time encoding,

using a hidden Markov model to determine scores indicating how the input speech matches some or all of a plurality of speech elements,

determining which, if any, speech element best corresponds to the input speech using both the time encoded speech and the Markov scores, and

outputting the speech element, if any, so determined.

15. A computer program according to claim 14 wherein the time encoding step includes

identifying the intervals between the occurrences of the input crossing a given value, and quantising the lengths of the intervals,

identifying the number of complex zeroes of the input parameter, up to a predetermined rank, in the said intervals, and

recording the quantised lengths of the intervals and a measure of the said number complex zeroes up to a predetermined rank as a representation of the variation of the input parameter.