Speaker Recognition Using Auditory Features and Polynomial Classifier

Abstract

This paper presents a speaker recognition method which makesuse of auditory features and polynomial classifier for speakerrecognition. Auditory features based on an auditory peripherymodel extract significant speaker characteristics. Polynomialclassifier has been used to accomplish speaker recognition task.Polynomial classifier has several advantages over theconventional classifiers such as computational scalability with thenumber of speakers, discriminative training allowing it to use outof class data and the statistical interpretation of scoring allowing it to combine with HMM and GMM. This approach achievessubstantial performance improvement in a speaker identificationtask compared with state-of-the-art in a wide range of signal tonoise conditions.