Text-independent speaker identification system based on the histogram of DCT-cepstrum coefficients

Abstract

There are several known feature sets for text-independent speaker-identification systems, most of which depend on spectral information. Among these feature sets as a most successful one, there is the set of the Mel-Frequency Cepstrum Coefficients (MFCC).

This paper introduces a new feature set, namely, the Histogram of the DCT-Cepstrum Coefficients, inspired by the common use of the MFCC, but simpler and faster in computation. A text-independent speaker-identification system based on the DCT-Cepstrum Histogram and Gaussian Mixture Model (GMM) is implemented. The new feature was tested using speech files from the ELSDSR database and TIMIT corpus. The new feature set managed to achieve high efficiency rates with speaker identification accuracy of 100% on 23 speakers from the ELSDSR database, and 99% on 630 speakers from the TIMIT corpus.