The Segmentation of Arabic word signal based on Eigenvalues and Eigenvectors principles

Abstract

In this paper, we use the principles of eigenvalues and eigenvectors for the first time in the field of Arabic word segmentation. In the beginning, the segmentation algorithm calculates eigenvalues and eigenvectors for each input speech signal, and selects the largest values. The values of final vector are used for linear separation to consonant regions and vowel regions by using threshold value which calculated from the final vector itself. The segmentation points are extracted efficiently and determine the beginning and the end points of each phoneme of the input word signal in high accuracy. Experimental results for a number of isolated Arabic words for a number of speakers ( males and females) given in the paper show that the algorithm determines the beginning and the end points for each phoneme in the input speech signals efficiently.