Speaker recognition systems

Abstract

Speaker recognition (identification and / or verification) methods and systems, in which speech models for enrolled speakers consist of sets of feature vectors representing the smoothed frequency spectrum of each of a plurality of frames and a clustering algorithm is applied to the feature vectors of the frames to obtain a reduced data set representing the original speech sample, and wherein the adjacent frames are overlapped by at least 80%. Speech models of this type model the static components of the speech sample and exhibit temporal independence. An identifier strategy is employed in which modelling and classification processes are selected to give a false rejection rate substantially equal to zero. Each enrolled speaker is associated with a cohort of a predetermined number of other enrolled speakers and a test sample is always matched with either the claimed identity or one of its associated cohort. This makes the overall error rate of the system dependent only on the false acceptance rate, which is determined by the cohort size. The false error rate is further reduced by use of multiple parallel modelling and / or classification processes. Speech models are normalised prior to classification using a normalisation model derived from either the test speech sample or one of the enrolled speaker samples (most preferably from the claimed identity enrolment sample).

Description

technical field [0001] The present invention relates to a system, method and apparatus for performing speaker recognition. Background technique [0002] Speaker recognition includes the related fields of speaker verification and speaker identification. Its main purpose is to confirm the speaker's claimed identity from his / her voice, which is called authentication, or to identify the speaker from his / her voice, which is called discrimination. Both of these use the human voice as a biometric measure and infer the unique relationship between a voice and the person who produced it. This unique relationship enables verification and discrimination. Speaker recognition techniques analyze test utterances and compare them to known templates or models for the person being identified or authenticated. The efficiency of this system depends on the quality of the algorithms used in the processing. [0003] Speaker recognition systems have a variety of applications. According to anoth...

AI Technical Summary

Problems solved by technology

Third, HMM models inherently do not allow efficient model-to-model comparisons

Therefore, the important structural details contained in the enrolled speaker data set cannot be efficiently analyzed and used to improve the performance of the system

Fourth, HMM technology uses temporal information to construct models, so it is vulnerable to pretenders who imitate other people's voices by temporarily changing pitch, etc.

Fifth, the required speaker cannot easily optimize the world model / fake cohort used by the system for the purpose of testing utterances

However, the mean of the true speaker distribution will be considerably lower than the mean of the impostor distribution tested with the same model

Images