Voice authentication system

• US 20080172230A1
• Filed: 08/17/2007
• Published: 07/17/2008
• Est. Priority Date: 02/18/2005
• Status: Active Grant

• Current Assignee: Fujitsu Limited
• Original Assignee: Fujitsu Limited
• Inventors: Hayakawa, Shoji
• Granted Patent: US 7,657,431 B2
• Time in Patent Office: Days
• US Class Current: 704/249
• CPC Class Codes: G10L 17/14 Use of phonemic categorisation or speech recognition prior to speaker recognition or verification

Abstract

A text-dependent voice authentication system that performs authentication by urging a user to input a keyword by voice includes: an input part (11) that receives a voice input of a keyword divided into a plurality of portions with an utterable unit being a minimum unit over a plurality of times at a time interval for each of the portions; registered speaker-specific syllable model DB (20) that previously stores a registered keyword of a user as a speaker model created in the utterable unit; a feature value conversion part (12) that obtains a feature value of a voice contained in a portion of the keyword received by the first voice input in the input part (11) from the portion; a similarity calculation part (13) that obtains a similarity between the feature value and the speaker model; a keyword checking part (17) that determines whether or not voice inputs of all the syllables or phonemes configuring an entire registered keyword by the plurality of times of voice inputs, based on the similarity obtained in the similarity calculation part; and an authentication determination part (19) that determines whether to accept or reject authentication, based on a determination result in the keyword checking part and the similarity obtained in the similarity calculation part.

Claims

1. A text-dependent voice authentication system that performs authentication by receiving a keyword which a user inputs by voice, comprising:

an input part that receives a voice input of a keyword from the user, said keyword is divided into a plurality of portions with an utterable unit which is a minimum unit, said voice input is carried out over a plurality of times at a time interval for each of the portions;

a speaker model storage part that previously stores a registered keyword of the user as a speaker model created in the utterable unit;

a feature value conversion part that obtains a feature value of a voice from a portion of the keyword which is received by a first voice input in the input part;

a similarity calculation part that obtains a similarity between the feature value obtained in the feature value conversion part and the speaker model;

a speech content determination part that determines information on a speech content by the plurality of times of voice inputs, based on the similarity obtained in the similarity calculation part;

a keyword checking part that determines whether or not the speech content of the plurality of times of voice inputs is capable of configuring an entire registered keyword, based on the information on the speech content determined in the speech content determination part; and

an authentication determination part that determines whether to accept or reject authentication, based on a determination result in the keyword checking part and the similarity obtained in the similarity calculation part.

2. The voice authentication system according to claim 1, wherein the utterable unit is a syllable.

3. The voice authentication system according to claim 2, wherein in the speaker model storage part, a discrete index in the speaker model storage part is provided to each speaker model corresponding to each syllable configuring the registered keyword,

the feature value conversion part obtains each feature value for each syllable from each portion of the keyword received by each voice input,

the similarity calculation part obtains each similarity between each feature value for each syllable and the speaker model corresponding to each syllable,

the system further comprises a syllable determination part that determines to which syllable of the registered keyword any of the portions of the keyword received by the voice inputs is the most similar, based on the similarity obtained in the similarity calculation part, and

the keyword checking part determines whether or not the syllables determined by the plurality of times of voice inputs are capable of configuring the entire registered keyword, based on a determination result of the syllable determination part.

4. The voice authentication system according to claim 1, wherein the utterable unit is a reading of a numeric or a reading of an alphabet.

5. The voice authentication system according to claim 1, further comprising a speech recognition part that recognizes a syllable or phoneme of the portion of the keyword using a speaker independent speech recognition method, from the feature value obtained in the feature value conversion part,

wherein the keyword confirmation part determines whether or not it is possible to configure the entire registered keyword using a result of speech recognition by the plurality of times of voice inputs in the speech recognition part.

6. The voice authentication system according to claim 1, wherein

the authentication determination part rejects authentication in a case where both there is no voice input of a subsequent portion even after an elapse of a predetermined time from a completion of the voice input of the portion of the keyword and it is not possible to configure the entire registered keyword, using information on the speech content of the voice inputs up to that time.

7. The voice authentication system according to claim 1, further comprising:

a positional information acquiring part that acquires location information of the user every time the portion of the keyword is input by voice; and

a position checking part that compares location information acquired in the positional information acquiring part at a time of a previous voice input with location information acquired in the positional information acquiring part at a time of a current voice input and checks whether or not the user has moved by a predetermined distance or longer from the previous voice input to the current voice input using a result of the comparison.

8. The voice authentication system according to claim 1, further comprising a similarity integration part that obtains an integrated similarity by integrating similarities obtained in the similarity calculation part, regarding all the portions of the keyword received by the plurality of times of voice inputs,

wherein the authentication determination part determines whether to accept or reject authentication on the basis of the integrated similarity obtained in the similarity integration part.

9. The voice authentication system according to claim 1, wherein the input part receives a voice input through a mobile terminal of the user.

10. A computer program product stored on a computer-readable medium, for causing a computer to embody a text-dependent voice authentication system that performs authentication by receiving a keyword which a user to input a keyword by voice, said computer program comprising the operations of

an input operation of receiving a voice input of a keyword from the user, said keyword is divided into a plurality of portions with an utterable unit which is a minimum unit, said voice input is carried out over a plurality of times at a time interval for each of the portions;

a feature value conversion operation of obtaining a feature value of from a portion of the keyword which is received by a first voice input in the input part, said feature value of a voice is a speech signal contained in the portion of the keyword;

a similarity calculation operation of referring to a speaker model storage part in which a keyword of a user is previously registered as a speaker model created in the utterable unit, and obtaining a similarity between the feature value obtained in the feature value conversion operation and the speaker model;

a speech content determination operation of determining information on a speech content by the plurality of times of voice inputs, based on the similarity obtained in the similarity calculation part;

a keyword checking operation of determining whether or not the speech content of the plurality of times of voice inputs is capable of configuring an entire registered keyword, based on the information on the speech content determined in the speech content determination operation; and

an authentication determination operation of determining whether to accept or reject authentication, based on a determination result by the keyword checking operation and the similarity obtained by the similarity calculation operation.