Text independent speaker recognition for transparent command ambiguity resolution and continuous access control

• US 6,073,101 A
• Filed: 01/28/1997
• Issued: 06/06/2000
• Est. Priority Date: 02/02/1996
• Status: Expired due to Fees

• Current Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Original Assignee: International Business Machines Corporation
• Inventors: Maes, Stephane Herman
• Primary Examiner(s): Hudspeth, David R.
• Assistant Examiner(s): Smits, Talivaldis Ivars
• Application Number:
  US08/788,471
• Time in Patent Office: 1,225 Days
• Field of Search: 704/231, 704/246, 704/275
• US Class Current: 704/275
• CPC Class Codes: G10L 15/065 Adaptation
• CPC Class Codes: G10L 17/04 Training, enrolment or model building
• CPC Class Codes: G10L 17/14 Use of phonemic categorisation or speech recognition prior to speaker recognition or verification

Abstract

Feature vectors representing each of a plurality of overlapping frames of an arbitrary, text independent speech signal are computed and compared to vector parameters and variances stored as codewords in one or more codebooks corresponding to each of one or more enrolled users to provide speaker dependent information for speech recognition and/or ambiguity resolution. Other information such as aliases and preferences of each enrolled user may also be enrolled and stored, for example, in a database. Correspondence of the feature vectors may be ranked by closeness of correspondence to a codeword entry and the number of frames corresponding to each codebook are accumulated or counted to identify a potential enrolled speaker. The differences between the parameters of the feature vectors and codewords in the codebooks can be used to identify a new speaker and an enrollment procedure can be initiated. Continuous authorization and access control can be carried out based on any utterance either by verification of the authorization of a speaker of a recognized command or comparison with authorized commands for the recognized speaker. Text independence also permits coherence checks to be carried out for commands to validate the recognition process.

Claims

1. A method of performing text independent speaker recognition including speaker identification, comprising:

sampling overlapping frames of a speech signal,

computing a feature vector for each said frame of said speech signal,

comparing each said feature vector with vector parameters and variances stored in a codebook corresponding to an enrolled speaker,

accumulating a number of frames for which the corresponding feature vector corresponds to vector parameters and variances in a codebook,

identifying an enrolled speaker or detecting a new speaker in response to results of said accumulating step or said comparing step, respectively, controlling a function in accordance with said identifying step, and

initiating an enrollment procedure when a new speaker is detected before or during control of said function.

2. A method as recited in claim 1, wherein said enrollment procedure includes the step of

presenting an enrollment menu.

3. A method of performing text independent speaker recognition comprising:

sampling overlapping frames of a speech signal;

computing a feature vector for each said frame of said speech signal;

comparing each said feature vector with vector parameters and variances stored in a codebook corresponding to an enrolled speaker;

accumulating a number of frames for which the corresponding feature vector corresponds to vector parameters and variances in a codebook;

identifying an enrolled speaker or detecting a new speaker in response to results of said accumulating step or said comparing step, respectively,

recognizing a command or a plurality of commands within said input speech signal;

retrieving enrolled information corresponding to said speaker identified in said identifying step; and

interpreting said command in accordance with said enrolled information retrieved in said retrieving step,

wherein said command is carried out by a procedure which differs between enrolled speakers, said method including the further step of;

selecting a procedure to carry out said command in accordance with said enrolled information retrieved by said retrieving step.

4. A method as recited in claim 3, wherein a command of said plurality of commands is selected in accordance with said enrolled information retrieved by said retrieving step.

5. A method as recited in claim 3, including the further step of

providing feedback of a result of said interpreting step to said speaker.

6. A method as recited in claim 3, including the further step of

verifying the speaker of said command recognized by said step of recognizing a command.

7. A method as recited in claim 3, including the further steps of

retrieving a command which is authorized for a speaker identified by said identifying step, and

comparing said command which is authorized for said speaker with said command recognized by said recognizing step.

8. A method as recited in claim 7, wherein said step of retrieving a command includes the step of accessing a look-up table in accordance with said speaker identified by said identifying step.

9. A method as recited in claim 3, including the further step of

checking coherence of said command recognized by said recognizing step.

10. A method of performing text independent speaker recognition including speaker identification and verification, comprising the steps of:

sampling overlapping frames of a speech signal,

computing a feature vector for each said frame of said speech signal,

comparing each said feature vector with vector parameters and variances stored in a codebook corresponding to an enrolled speaker,

accumulating a number of frames for which the corresponding feature vector corresponds to vector parameters and variances in a codebook,

identifying an enrolled speaker or detecting a new speaker in response to results of said accumulating step or said comparing step, respectively,

controlling a function in accordance with said identifying step, and

verifying the speaker identified in said identifying step in accordance with steps that include;

performing consistency checks;

building a cohort of similar speakers; and

comparing the speaker counts to the cohort counts during the identification.

11. Apparatus including a speech recognition system and a text independent speaker recognition system, said text independent speaker recognition system comprising:

means for sampling overlapping frames of a speech signal,

means for computing a feature vector for each said frame of said speech signal,

means for identifying a speaker and controlling a function including one of;

(a) means for comparing each said feature vector with vector parameters and variances stored in a codebook corresponding to an enrolled speaker, and

(b) means for accumulating in accordance with respective codebooks, a number of frames for which feature vectors correspond to vector parameters and variances in a codebook corresponding to a respective enrolled speaker,

means for recognizing a command or a plurality of commands,

means for retrieving enrolled information corresponding to said speaker identified by said identifying means, and

means for interpreting said command in response to said means for retrieving said enrolled information,

wherein said command is carried out by a procedure which differs between enrolled speakers, said apparatus further including;

means for selecting a procedure to carry out said command in accordance with said enrolled information.