Method of accommodating for carbon/electret telephone set variability in automatic speaker verification

• US 5,528,731 A
• Filed: 11/19/1993
• Issued: 06/18/1996
• Est. Priority Date: 11/19/1993
• Status: Expired due to Term

• Current Assignee: Avaya Incorporated
• Original Assignee: AT&T Corporation (AT&T, Inc.)
• Inventors: Schoeffler, Max S., Sachs, Richard M.
• Primary Examiner(s): MacDonald, Allen R.
• Assistant Examiner(s): ONKA, THOMAS
• Application Number:
  US08/155,973
• Time in Patent Office: 942 Days
• Field of Search: 395/2.42, 395/2.55, 381/103, 381/98
• US Class Current: 704/246
• CPC Class Codes: G10L 17/20 Pattern transformations or operations aimed at increasing system robustness, e.g. against channel noise or different working conditions

Abstract

In a speaker verification system, a method of compensating for differences in speech samples obtained during registration and those obtained during verification due to the use of different types of microphones is provided by filtering at least one of the samples such that the similarities of the two samples are increased. The filtered sample is used within the speaker verification matching process. A two-way comparison is disclosed in which both a verification speech sample and a reference sample are filtered with nonlinear microphone characteristics such as carbon microphone characteristics. A four-way comparison is also disclosed in which patterns produced from unfiltered verification and reference samples and patterns produced from the filtered verification and reference samples are compared to identify a match. A score is determined for each comparison. The comparison having the best score is used to determine if a match has occurred.

Claims

1. A method of processing speech samples for use in verifying a speaker, the method comprising the steps of:

receiving speech samples generated by different types of microphones;

filtering at least one of said speech samples generated by one of the microphone types with the characteristics of another one of the microphone types, wherein the characteristics of said another one of the microphone types are non-linear; and

producing patterns from said filtered speech samples that can be used to verify the speaker.

2. The method according to claim 1 wherein said speech samples include reference speech samples.

3. The method according to claim 1 wherein said speech samples include verification speech samples.

4. The method according to claim 1 wherein said another one of the microphone types is a carbon microphone.

5. The method according to claim 1 wherein the characteristics of said one microphone type are linear.

6. The method according to claim 5 wherein said microphone is an electret microphone.

7. A method of verifying a speaker, the method comprising the steps of:

receiving a speech sample provided for verification from one type of microphone;

filtering the verification speech sample with the characteristics of another microphone type, wherein the characteristics of said another microphone type are non-linear;

producing a verification pattern from the filtered verification speech sample;

comparing the verification pattern to a generated reference pattern produced from a reference speech sample provided during a registration session; and

verifying the identity of the speaker as a function of the reference pattern and verification pattern.

8. The method according to claim 7 wherein said generated reference pattern is produced by filtering the reference speech sample with the characteristics of said another microphone type.

9. The method according to claim 7 wherein said generated reference pattern comprises a filtered reference pattern produced by filtering the reference speech sample with the characteristics of said another microphone type, the method further comprising the steps of:

producing an unfiltered verification pattern from a verification speech sample which has not been filtered;

producing an unfiltered reference pattern from a reference speech sample which has not been filtered; and

comparing each verification pattern with each reference pattern.

10. The method according to claim 9 wherein said determining step further comprises the steps of:

calculating a score representing the similarity of the verification pattern and the reference pattern for each comparison;

determining which comparison has the best score based on predetermined criteria.

11. A method according to claim 7 wherein said another microphone type is a carbon microphone.