Effective arabic speech segmentation strategy

Abstract

Speech segmentation is a process to segment speech utterances into small chunks, where each chunk represents a phoneme. The phoneme is an essential unit in any speech which is recognizable. In this paper, a segmentation speech approach was proposed to segment consonant and vowel phonemes from speech utterances of Arabic basic syllables, in order to analyze the consonant production of a group of Malay-speaking normal hearing children and adults. The approach is a combination of zero-crossing counts and signal energy. The zero-crossing counts were used to extract the noise signals, whilst the signal energy was utilized for identifying the speech signals. The spectrogram was used to determine the frequencies with the most intense energy.