Speaker identification using radial basis function neural network

Abstract EN

The main objective of this paper is to apply radial basis function neural networks (RBFNN) and evaluated its performance by comparing the results with other method.

In this paper two feature vectors are used separately to address speaker identification problem.

The features are linear predictive code (LPC) and Mel-frequency cepstral coefficient (MFCC).

The radial basis function neural network (RBFNN) approach is used for matching purpose.

This work proposes can be summarized into three steps.

The first step is to frame and windowing the input speech signal using hamming window.

The second step is to extract the reference and test speech signal using LPC or MFCC as feature extraction.

Finally, in the third step, radial basis function neural network has been used to perform the similarity between the test and reference templates.

The results show that speaker identification using MFCC and RBFNN gives (100 %) identification rate and higher identification rate compared with other method.