This paper presents the initial effort to perform speaker verification by utilizing the speech signal characteristics found in individual’s voice to recognize its speaker. A total of six speakers from different backgrounds were selected as sample and each of them is required to pronounces numbers zero to nine for 5 times. The recorded speech signal then undergoes a series of speech processing, which contains Pre-emphasis, Framing, Windowing and Endpoint Detection. To obtain the features of each speech signal, the Linear Prediction Coefficients (LPC) technique is used. The collection of LPC coefficients then were feed to the Multilayer Perceptron Neural Network trained by Back Propagation algorithm, which acts as a pattern matching algorithm. The results show that the speech signal has the potential to be used to verify its speaker in high accuracy.
