SPEAKER IDENTIFICATION USING HYBRID MODEL OF PROBABILISTIC NEURAL NETWORK AND FUZZY C-MEANS

Abstract

A hybrid model of Probabilistic Neural Network and Fuzzy C-Means has been developed. The model has been applied using Mel Frequency Cepstrum Coefficients (MFCC) as feature extraction for identification. In addition to the natural voice, the effect of noise has also been taken into account. It has been shown that the model has good accuracy at 96% for voice without noise, 85.5% for voice with noise at the level of signal to noise ratio 30, and 60% for voice with noise at the level of signal to noise ratio 20. It has also been concluded that the clustering procedure using Fuzzy C-Means could improve the accuracy up to 96% for large number of training data.