On the speaker verification using the TESPAR coding method

Abstract

This work describes study on the speaker verification rate, using the TESPAR (Time Encoding Signal Processing and Recognition) coding method, when the speech signal is sampled at different rates. The effect of filtering on the speech signal was studied, as well. The TESPAR method is a processing and recognition method in the time domain, proposed by R.A. King. The key problem is to define the TESPAR alphabet used for the TESPAR coding process. In this paper is proposed an approach to generate this alphabet using the Kohonen Neural Networks in a vector quantization process. For the recognition process parallel Multi Layer Perceptron (MLP) neural network were used. As inputs for training/test vectors of the MLP-NN, the TESPAR-S matrices were employed.