Voice processing and verification system

• US 6,233,556 B1
• Filed: 12/16/1998
• Issued: 05/15/2001
• Est. Priority Date: 12/16/1998
• Status: Expired due to Term

• Current Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Original Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Inventors: Shahshahani, Ben, Teunen, Remco
• Primary Examiner(s): Dorvil, Richemond
• Application Number:
  US09/216,622
• Time in Patent Office: 881 Days
• Field of Search: 704/200, 704/201, 704/231, 704/241, 704/246, 704/250, 704/270, 704/243, 704/255, 704/256
• US Class Current: 704/250
• 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 17/20 Pattern transformations or operations aimed at increasing system robustness, e.g. against channel noise or different working conditions
• CPC Class Codes: G10L 21/0216 characterised by the method used for estimating noise

Abstract

A voice processing and verification system accounts for variations dependent upon telephony equipment differences. Models are developed for the various types of telephony equipment from many users speaking on each of the types of equipment. A transformation algorithm is determined for making a transformation between each of the various types of equipment to each of the others. In other words, a model is formed for carbon button telephony equipment from many users. Similarly, a model is formed for electret telephony equipment from many users, and for cellular telephony equipment from many users. During an enrollment, a user speaks to the system. The system forms and stores a model of the user'"'"'s speech. The type of telephony equipment used in the original enrollment session is also detected and stored along with the enrollment voice model. The system determines the types of telephony equipment being used based upon the spectrum of sound it receives. The telephony equipment type determination is based upon models formed for each of the telephony equipment types spoken by many different users. Thereafter, when a current user calls in, his/her voice will be compared to the stored model if the same telephony equipment as used in the enrollment is determined. If the user calls in on another type of equipment than that used during the enrollment, the transformation for telephony equipment is applied to the model. The user'"'"'s voice is then verified against the transformed model. This improves the error rate resulting from different telephony equipment types.

Claims

1. A method of voice processing to verify a user comprising the steps of:

a. receiving an incoming user speech;

b. determining the particular type of equipment corresponding to the incoming user speech;

c. retrieving an enrollment speech model;

d. determining the particular type of equipment corresponding to the enrollment speech model; and

e. transforming the enrollment speech model to a corresponding equipment type for the incoming user speech.

2. The method of voice processing to verify a user according to claim 1 further comprising the step of generating a plurality of speaker independent models each speaker independent model representing a particular type of equipment, such that the plurality of speaker independent models are used in the step of transforming.

3. The method of voice processing to verify a user according to claim 1 further comprising the step of comparing the incoming user speech to the transformed speech model to verify the user.

4. The method of voice processing to verify a user according to claim 1 wherein the model is a gaussian mixture model.

5. A method of voice processing to verify a user comprising the steps of:

a. receiving an incoming user speech and forming a model thereof to form an incoming speech model;

b. determining the particular type of equipment corresponding to the incoming user speech;

c. retrieving an enrollment speech model;

d. determining the particular type of equipment corresponding to the enrollment speech model; and

e. transforming the incoming speech model to a corresponding equipment type for the enrollment speech model.

6. The method of voice processing to verify a user according to claim 5 further comprising the step of generating a plurality of speaker independent models each speaker independent model representing a particular type of equipment, such that the plurality of speaker independent models are used in the step of transforming.

7. The method of voice processing to verify a user according to claim 5 further comprising the step of comparing the incoming speech model to the transformed speech model to verify the user.

8. The method of voice processing to verify a user according to claim 5 wherein the models are gaussian mixture models.

9. A method of voice processing to verify a user comprising the steps of:

a. receiving an incoming user speech;

b. determining the particular type of equipment corresponding to the incoming user speech;

c. retrieving an enrollment speech sample;

d. determining the particular type of equipment corresponding to the enrollment speech sample;

e. forming a model from the enrollment speech sample to form an enrollment speech model; and

f. transforming the enrollment speech model to a corresponding equipment type for the incoming user speech.

10. The method of voice processing to verify a user according to claim 9 further comprising the step of generating a plurality of speaker independent models each speaker independent model representing a particular type of equipment, such that the plurality of speaker independent models are used in the step of transforming.

11. The method of voice processing to verify a user according to claim 9 further comprising the step of comparing the incoming user speech to the transformed enrollment speech model to verify the user.

12. The method of voice processing to verify a user according to claim 9 wherein the model is a gaussian mixture model.

13. A method of voice processing to verify a user comprising the steps of:

a. receiving an incoming user speech;

b. determining the particular type of equipment corresponding to the incoming user speech;

c. forming a model from the incoming user speech to form an incoming user speech model;

d. retrieving an enrollment speech sample;

e. determining the particular type of equipment corresponding to the enrollment speech sample; and

f. transforming the incoming user speech model to a corresponding equipment type for the enrollment speech sample.

14. The method of voice processing to verify a user according to claim 13 further comprising the step of generating a plurality of speaker independent models each speaker independent model representing a particular type of equipment, such that the plurality of speaker independent models are used in the step of transforming.

15. The method of voice processing to verify a user according to claim 13 further comprising the step of comparing the transformed incoming user speech model to the enrollment speech sample to verify the user.

16. The method of voice processing to verify a user according to claim 13 wherein the model is a gaussian mixture model.

17. A method of voice processing to identify a user comprising the steps of:

a. receiving an incoming voice sample;

b. digitally processing the incoming voice sample thereby creating a digitally processed voice sample;

c. storing the digitally processed voice sample in a buffer;

d. retrieving an enrollment voice sample and generating a model of the enrollment voice sample;

e. comparing the model of the enrollment voice sample to the digitally processed voice sample, in order to determine a match within a certain tolerance.

18. The method of voice processing and verification, as in claim 17, wherein the step of receiving an incoming voice sample is done by recording an incoming user voice signal over a fixed period of time.

19. The method of voice processing and verification, as in claim 17, wherein the enrollment sample is recorded, digitally processed and stored in a memory device before receiving the incoming voice sample.

20. The method of voice processing and verification, as in claim 19, wherein the model of the enrollment sample is formed before receiving the incoming voice sample.

21. The method of voice processing and verification, as recited in claim 19, further comprising the step of identifying the type of equipment used in recording the enrollment sample and storing this information in the memory device.

22. The method of voice processing and verification, as recited in claim 17, wherein the plurality of models of the enrollment voice sample are generated by impressing transformations on the enrollment voice sample.

23. The method of voice processing and verification, as in claim 22, wherein the transformations impressed on the enrollment voice sample represent differences attributable to alternate recording equipment.

24. The method of voice processing and verification, as recited in claim 22, wherein the transformations are programmable.

25. A method of voice processing to verify a user comprising the steps of:

a. receiving an enrollment voice sample;

b. digitally processing the enrollment voice sample thereby creating a digitally processed enrollment sample;

c. storing the digitally processed enrollment sample in a memory device;

d. retrieving an incoming voice signal and generating a plurality of signal models of said incoming voice signal;

e. comparing each of the signal models of the incoming voice signal to the digitally processed enrollment sample to determine a match within a certain system tolerance.

26. The method of voice processing to verify a user, as in claim 25, wherein the step of generating a plurality of signal models of said incoming voice signal is accomplished by impressing transformations on the incoming voice signal.

27. The method of voice processing to verify a user, as in claim 26, wherein the transformations impressed on the incoming voice signal represent differences attributable to alternate recording equipment.

28. The method of voice processing to verify a user, as recited in claim 26, wherein the transformations are programmable.

29. A computer apparatus for telephonic voice processing and verification, designed to minimize false rejection or false acceptance, consisting of:

a. means for receiving a telephonic voice sample over a fixed period of time;

b. means for digitally processing the telephonic voice sample;

c. means for storing the digitally processed telephonic voice sample; and

d. means for retrieving the digitally processed voice sample and generating a plurality of models of the digitally processed voice sample; and

e. means for comparing the plurality of models of the digitally processed voice sample with an enrollment signal, in order to determine a match within certain known variable tolerances.

30. The computer apparatus for telephonic voice processing and verification, as in claim 29, further comprising:

a. means for recording an enrollment signal;

b. means for digitally processing the enrollment signal; and

c. means for storing the enrollment signal in a memory device.

31. The computer apparatus for telephonic voice processing and verification, as in claim 30, further comprising a means for identifying the type of telephony equipment used in recording the enrollment signal.

32. The computer apparatus for telephonic voice processing and verification, as in claim 29, wherein the certain variable tolerances are programmable.

33. A computer apparatus for voice processing and verification, designed to minimize false rejection or false acceptance, comprising:

a. means for receiving an enrollment voice sample;

b. means for digitally processing the enrollment voice sample thereby creating a digitally processed enrollment sample;

c. means for storing the digitally processed enrollment sample;

d. means for retrieving the digitally processed enrollment sample and generating a plurality of signal models of said digitally processed enrollment sample; and

e. means for comparing each of the signal models of the digitally processed enrollment sample to an incoming voice sample to determine a match within a certain system tolerance.

34. The computer apparatus for voice processing and verification, as in claim 33, wherein the means for generating a plurality of signal models of said digitally processed enrollment sample includes a means for impressing transformations on the digitally processed enrollment sample.

35. The computer apparatus for voice processing and verification, as in claim 34, wherein the transformations impressed on the digitally processed enrollment sample represent differences attributable to alternate recording equipment.

36. The computer apparatus for voice processing and verification, as recited in claim 35, wherein the transformations are programmable.

37. A voice processing and verification system for verifying a user comprising:

a. one or more telephone lines coupled to a telephone call detection system;

b. a voice processing module coupled to the telephone lines wherein the voice processing module includes a digital voice recording unit for recording incoming telephone calls from both new users and current users to the system, and a voice processing server for digitally processing these recordings;

c. a common memory coupled to the voice processing module for storing the digitally processed recording of a new user to the system, known as an enrollment sample;

d. a buffer for storing the digitally processed recording of a current user to the system;

e. a voice modulating unit coupled to the common memory for retrieving the enrollment sample and generating a plurality of various models of the enrollment sample, as if the enrollment sample were originally recorded over various types of telephony equipment; and

f. a voice comparison module for comparing each of the plurality of various models of the enrollment sample with the digitally processed recordings of a current user to the system stored in the buffer in order to determine a match within a certain threshold.

38. The voice processing and verification system, as in claim 37, wherein the voice modulating unit generates the plurality of signal models of said enrollment sample by impressing transformations on the enrollment sample.

39. The voice processing and verification system, as in claim 38, wherein the transformations impressed on the enrollment sample represent differences attributable to alternate recording equipment.

40. The voice processing and verification system, as recited in claim 39, wherein the transformations are programmable.