Speaker Verification With The Constraint of Limited Data

INTRODUCTION

¢ Speaker Recognition is the identification of the person who is speaking by characteristics of their voices.

¢ Speaker recognition is basically divided into two categories:

Ø Speaker Identification

Ø Speaker Verification

¢ Speaker recognition can be characterized into:

Ø Text Dependent

Ø Text Independent

¢ Enrolment phase: client provides speech to build speaker model

¢ Verification phase:

compare caller’s speech with client’s model

if it matches well enough, accept caller as being the client (assuming other security checks OK)

otherwise reject caller’s identity claim

¢ The system operates in two modes: A training mode & a Recognition mode.

¢ The method for modelling the human auditory perception system, Mel Frequency Cepstral Coefficients.

¢ Universal Background model algorithms are used to implement the speaker.

By Manjunath B