Text independent speaker identification using wavelet transform
Gyanendra K. Verma, Uma Shanker Tiwary · 2010
The purpose of this paper is to evolve a robust text independent speaker identification system based on the wavelet transform, which is able to analyze signal at multiple resolutions. The proposed system identifies speakers by their acoustic characteristics embedded in speech signal of speakers. Features are obtained from approximation and detail coefficients by calculating entropy, standard deviation and mean at each decomposition level. The similarity between the extracted features and a set of reference features is calculated by means of K-NN classifier to determine the speaker's identity. The results are evaluated with VoxForge speech corpus containing more than 200 speakers and each speaker profile contains ten speech signals. The results achieved are promising with 83.9 % accuracy for 200 speakers. Some popular existing methods are also evaluated for comparison in this paper. The result shows that the proposed method is more effective and robust than other methods. The performance of our system is very good with noisy dataset at different SNR.