Source and system features for text independent speaker identification using iterative clustering approach

Abstract

The main objective of this paper is to explore the effectiveness of perceptual features combined with pitch for text independent speaker recognition. The proposed combined features are captured and training models are developed by K-means clustering procedure. Speaker recognition system is evaluated on clean test speeches and the experimental results reveal the performance of the proposed algorithm in performing speaker recognition based on minimum distance between test features and clusters. This algorithm gives the overall accuracy of 99.675% and 98.75% for the combined features and perceptual features respectively for identifying speaker among 8 speakers chosen randomly from 8 different dialect regions in “TIMIT” database. It also gives average accuracy of 96.375% and 95.625% for perceptual linear predictive cepstrum combined with pitch and perceptual linear predictive cepstrum respectively for 8 speakers chosen randomly from the same dialect region. The noteworthy feature of speaker identification algorithm is to evaluate the testing procedure on identical messages for all speakers. In this work, Fratio is computed as a theoretical measure to validate the experimental results on speaker recognition.