Browse Prior Art Database

Method And System For User Authentication Using Homophone Name Sets

IP.com Disclosure Number: IPCOM000124637D
Original Publication Date: 2005-May-03
Included in the Prior Art Database: 2005-May-03
Document File: 3 page(s) / 29K

Publishing Venue

IBM

Abstract

Voice authentication is a well-known mechanism to establish that a user is who they say they are, whereby a user's voice is compared to his/her previously stored voice "print". Currently, a user has to provide the account number /user ID (to establish identity) and eventually also answer a challenge question (stored voice print phrase/name/randomly generated number) to perform the voice authentication. The current authentication methods require the knowledge of the account number. The idea of this disclosure is to use only a user's name as identity claim and authentication mechanism, with the help of homophone sets. The advantage is that the user's account is not revealed (in public environments this step is giving away a user's private information), and the user does not need to remember or carry any account number information (credit cards, ID's etc).

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 54% of the total text.

Page 1 of 3

Method And System For User Authentication Using Homophone Name Sets

Voice authentication cannot be performed against an entire population (inefficient and unreliable). Currently, voice authentication mechanism is based on first subsetting the population (based on a user's ID/account number, which establishes a set of size one) and second, performing the actual voice authentication on the subset (this is a 2-step process).

Considering only a user's name, the resulting subset is at least a size of one, but could be greater than one because more people could have the exact same name, people could have nicknames identical to the user's name, or other names could be homophone of the user's name). However, the resulting subset is less than the population, thereby making voice print authentication feasible.

During an enrollment phase, with their identity already established, users speak their names. The voice print of their names will be stored in a database (the storing of voice prints can be performed with additional security of storing an obfuscated audio file, using a one way encryption mechanism such as described in U.S. Patent 5677989 - "Speaker verification system and process").

For speech recognition and performance purposes, grammars are generated apriori, containing all the names of the population (if the population size is too large, the user's location and name is required, in effect performing a double segmentation, first on location and second on name). When trying to access their account users are prompted to say their name. The recognition process, using the apriori generated grammar, returns the small "homophone set" of people containing identical names, nicknames identical to the user's name and homophones similar to the user's name (the set could be of size greater or equal to one). The successful returning of a set matching the name spoken by the user establishes the identity set of a user.

1

Page 2 of 3

Initialize Loop Counter: N=1

Note: This and the following steps are prior art and are not further represented

N>3?

Yes

Collect user's ID/Account Number or Transfer to Operator

No

Collect user's name

Increase Loop Counter: N=N+1

   Recognize name (ge...