Data processing: speech signal processing – linguistics – language – Speech signal processing – Recognition
Reexamination Certificate
1999-05-07
2001-10-23
Tsang, Fan (Department: 2648)
Data processing: speech signal processing, linguistics, language
Speech signal processing
Recognition
C704S246000
Reexamination Certificate
active
06308153
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates generally to voice verification and more particularly, to a voice verification system that verifies the identity of an individual based on voice samples collected during a telephone conversation.
2. Description of the Prior Art
Voice verification is needed in a variety of systems such as home banking, home inceration, remote database access, ticketless air travel etc. A common requirement of these systems is the need to verify an authorized user's identity who is trying to conduct a transaction at a remote location. Such a requirement is necessary in order to prevent an unauthorized user from gaining access who potentially can cause damage. The danger of an unauthorized user gaining access is especially high in today's computer literate society.
Other types of identification methods have proved to be limited or ineffective in such systems. For example, the use of passwords is limited by the fact that passwords may be forgotten, stolen or voluntarily given to another person. Other methods such as fingerprints, retinal scans etc. are inappropriate for remote transactions because the physical presence of the user to be identified is required. In contrast, voice verification systems provide a means to identify a potential user located anywhere within a telephone network.
Voice verification systems generally operate by comparing speech spoken by a potential user to previously stored speech containing corresponding words in order to identify the user. Usually the previously stored speech is entered into the system by an enrollment function. In a number of systems, the comparison between the spoken and stored speech is based on a measurement of the nearest neighbor distance between corresponding elements. This measurement is usually performed by computer processing of such elements converted into digital form.
An example of a voice verification system is exemplified by U.S. Pat. No. 5,339,385 to Higgins, entitled SPEAKER VERIFIER USING NEAREST NEIGHBOR DISTANCE MEASURE, issued on Aug. 16, 1994. Higgins discloses a system that includes a verification module that computes the nearest neighbor distance between a test session and an enrollment session. Higgins further discloses the verification module computing the nearest neighbor distances between the test session and a plurality of additional enrollment sessions from a group of reference speakers. The additional nearest neighbor distances are computed in order to minimize the probability of false acceptance.
Other examples of voice verification systems are exemplified by U.S. Pat. No. 5,271,088 to Bahler, entitled AUTOMATED SORTING OF VOICE MESSAGE THROUGH SPEAKER SPOTTING, issued on Dec. 14 1993 and U.S. application Ser. No. 08/510,321 to Naylor et al. Bahler discloses a system incorporating pre-processing techniques such as feature extraction and blind de-convolution, while Naylor et al discloses a system including a word recognizer utilizing Hidden Markov Modeling and a Viterbi Decoder.
Existing voice verification systems have a number of limitations. One limitation relates to the length of time required to enroll or verify a user into these systems. Very often the length of time is to long, which makes the use of these systems inconvenient or unacceptable to many users. Another limitation relates to the accuracy of the existing systems. The accuracy often is poor, due to the use of different phonesets for verification and enrollment.
Therefore, it as an object of the present invention to provide a voice verification system that reduces the amount of time required for the enrollment and verification.
Therefore, it as a further object of the present invention to provide a voice verification system that is accurate even though different phonesets are used for verification and enrollment.
SUMMARY OF THE INVENTION
A system and a method is disclosed for verifying a voice of a user prior to conducting a telephone transaction. The system and method includes a means for prompting the user to speak in a limited vocabulary. A feature extractor converts the sampled speech signal to a plurality of speech frames. A pre-processor is coupled to the feature extractor for processing the plurality of speech frames to produce a plurality of processed frames. The processing includes frame selection, which eliminates each of the plurality of speech frames having an absence of words.
A Viterbi decoder is coupled to the feature extractor for assigning a label to each of the plurality of speech frames to produce a plurality of frame labels. The plurality of processed frames, combined with the associated frame labels constitutes a voice model. The voice model includes each of the plurality of frame labels that correspond to the number of the plurality of processed frames.
The system and method further includes means for measuring a nearest neighbor distance between the voice model produced from speech of an unknown person and a user voice model produced from an enrollment speech of an enrolled user. The nearest neighbor distance is calculated by only comparing the individual frames of the voice model and the claimant voice model that have the same label. A means is also included for accepting or rejecting the claimed identity based on a comparison of the voice model to the claimant's voice model, and a comparison of the voice model to a plurality of alternative voice models. The identity is accepted if the voice model matches the claimant's model better than the alternative models.
REFERENCES:
patent: 5121428 (1992-06-01), Uchiyama et al.
patent: 5159638 (1992-10-01), Naito et al.
patent: 5167004 (1992-11-01), Netsch et al.
patent: 5271088 (1993-12-01), Bahler
patent: 5295223 (1994-03-01), Saito
patent: 5339385 (1994-08-01), Higgins
patent: 5341456 (1994-08-01), DeJaco
patent: 5459814 (1995-10-01), Gupta et al.
patent: 5649055 (1997-07-01), Gupta et al.
patent: 5687287 (1997-11-01), Gandhi et al.
patent: 5719921 (1998-02-01), Vysotsky et al.
patent: 5765127 (1998-06-01), Nishiguchi et al.
patent: 5774849 (1998-06-01), Benyassine et al.
patent: 5809455 (1998-09-01), Nishiguchi et al.
patent: 5832063 (1998-11-01), Vysotsky et al.
patent: 5832429 (1998-11-01), Gammel et al.
patent: 5839103 (1998-11-01), Mammone et al.
patent: 5862519 (1999-01-01), Sharma et al.
Bahler Lawrence G.
Higgins Alan L.
Huang William Y.
Hale and Dorr LLP
ITT Defense, Inc.
Opsasnick Michael N.
Tsang Fan
LandOfFree
System for voice verification using matched frames does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with System for voice verification using matched frames, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and System for voice verification using matched frames will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2604327