Speech recognition system

• US 5,754,978 A
• Filed: 10/27/1995
• Issued: 05/19/1998
• Est. Priority Date: 10/27/1995
• Status: Expired due to Fees

• Current Assignee: Speech Systems of Colorado, Inc.
• Original Assignee: Speech Systems of Colorado, Inc.
• Inventors: Perez-Mendez, Ivan, Anikst, Mark T., Trawick, David J.
• Primary Examiner(s): Hudspeth, David R.
• Assistant Examiner(s): EDOUARD, PATRICK NESTOR
• Application Number:
  US08/549,278
• Time in Patent Office: 935 Days
• Field of Search: 395/2.47, 395/2.48, 395/2.6, 395/2.64, 395/2.55, 395/2.49
• US Class Current: 704/255
• CPC Class Codes: G10L 15/32 Multiple recognisers used in sequence or in parallel; Score combination systems therefor, e.g. voting systems

Abstract

An improved speech recognition system with at least two speech recognition engines which may or may not be identical. Each of the speech recognition engines supply a recognized-text output signal each of which are provided to a text comparator. The comparator compares the recognized-text output signals and accepts or rejects the text based on the degree of agreement between the output signals from each of the engines. Completely different types of speech recognition engines can be employed. Alternatively, identical speech recognition engines could be employed with a slight perturbation added to one of the engines or the input signal of one of the engines. Alternatively, the very same speech recognition engine could be used at two different points in time, with one receiving some type of perturbation.

Claims

1. An improved speech recognition system for analyzing an input signal to recognize the speech contained within the signal and output an accepted and recognized text signal, comprising:

a first speech recognition engine receptive of the input signal, performing speech recognition processing, and supplying a recognized-text output signal which represents in textual form the speech recognized in the input signal;

a second speech recognition engine receptive of the input signal, performing speech recognition processing, and supplying a recognized-text output signal which represents in textual form the speech recognized in the input signal; and

a comparator receptive of the recognized-text output signals from the first and second speech recognition engines and supplying the accepted and recognized text output signal based upon the comparison of the recognized-text output signals, the accepted and recognized text output signal representing in textual form the speech recognized in the input signal and accepted by the computer.

2. An improved speech recognition system as defined in claim 1, wherein the comparison of the recognized-text output signals includes accepting portions of the text when the signals corresponding to those portions of the text are identical and rejecting those portions of the text when the signals corresponding to those portions of the text are different.

3. An improved speech recognition system as defined in claim 1, wherein the first and second speech recognition engines are identical except for a perturbation added to the second speech recognition engine.

4. An improved speech recognition system as defined in claim 3, wherein the perturbation is an engine perturbation which alters the processing of the second speech recognition engine.

5. An improved speech recognition system as defined in claim 3, wherein the perturbation is a signal perturbation which alters the input signal received and processed by the second speech recognition engine.

6. An improved speech recognition system as defined in claim 1, wherein the first and second speech recognition engines are of different types.

7. An improved speech recognition system as defined in claim 3, wherein the first and second speech recognition engines are the very same speech recognition engine performing the processing as the first and second engines at different points in time.

8. An improved speech recognition system as defined in claim 1, wherein the first and second speech recognition engines share a common first part of a speech recognition engine and have separate and distinct second parts of a speech recognition engine.

9. An improved speech recognition system as defined in claim 8, wherein part includes analog circuitry and the second part includes analog and digital circuitry.

10. An improved speech recognition system as defined in claim 8, wherein the first part includes analog circuitry and the second part includes digital circuitry.

11. An improved speech recognition system as defined in claim 1, wherein the recognized-text output signals from the first and second speech recognition engines include multiple possibilities of recognized text ranked in order of the calculated probability of being correct, and wherein the comparison includes comparing the ranked possibilities from the speech recognition engines.

12. An improved speech recognition system for analyzing an input signal to recognize the speech contained within the signal and output an accepted and recognized text signal, comprising:

multiple speech recognition engines each of which are receptive of the input signal and each of which supply a recognized-text output signal, each of the speech recognition engines being different from each other;

a comparator receptive of the recognized-text output signals from each of the speech recognition engines, the comparator supplying the accepted and recognized text output signal based upon the comparison of the recognized-text output signals.

13. An improved speech recognition system as defined in claim 12, wherein the comparison of the recognized-text output signals includes accepting portions of the text when the signals corresponding to those portions of the text are identical and rejecting those portions of the text when the signals corresponding to those portions of the text are different.

14. An improved speech recognition system as defined in claim 12, wherein the multiple speech recognition engines include three speech recognition engines and wherein the comparison of the recognized-text output signals includes accepting portions of the text when the signals from a majority of the speech recognition engines corresponding to those portions are identical and rejecting those portions of the text otherwise.

15. A method of recognizing speech represented in an input signal, comprising the steps of:

receiving the input signal;

creating a first intermediate signal representative of the input signal;

recognizing the speech contained in the first intermediate signal;

supplying a first recognized-text output signal representative of the speech recognized in the first intermediate signal;

creating a second intermediate signal representative of the input signal;

recognizing the speech contained in the second intermediate signal;

supplying a second recognized-text output signal representative of the speech recognized in the second intermediate signal;

comparing the first and second recognized-text output signals; and

accepting portions of the compared signals that are identical.

16. A method as defined in claim 15, further including the step of:

rejecting portions of the compared signals that are not identical.

17. A method as defined in claim 15, wherein the first and second intermediate signals are identical and the recognizing steps are different.

18. A method as defined in claim 15, wherein the first and second intermediate signals are different and the recognizing steps are identical.

19. A method as defined in claim 15, wherein the first and second intermediate signals are different and the recognizing steps are different.

20. A method as defined in claim 15, wherein the recognizing steps are different by adding a perturbation to one of the recognizing steps.

21. A method as defined in claim 15, wherein the first and second intermediate signals are different by adding a perturbation to one of the intermediate signals.

22. A method as defined in claim 15, wherein the creating steps are accomplished with analog circuitry.