Segmentation of Arabic letters signal using Multiscale Principal Component analysis and Zero-Crossing Rate based on Malay speakers

In this paper, an investigation of segmentation method for Arabic letters signals spoken by Malay speakers, which is the main step for further speech processing for the purpose of identifying proper pronunciation is performed will be discussed. Recording any corpus requires suitable environment in order to reduce the noise but in our application mobility is the most vital point for speech recording. Hence, in this study, Multiscale Principal Component is applied to de-noise the signal as a pre-step before employing Zero-Crossing Rate function that will further be utilized to segment the actual voice signal. The outcome of the proposed method will be the first phase for Arabic Speech Recognition under noise environment and uncontrolled conditions.