Abstract
This paper presents a proposed technique of Arabic speech phoneme recognition system; this system is useful in applications of Arabic speech processing. A propose technique is (Mutiridgelet transform) use for Arabic phoneme recognition. The procedure followed in this technique can be divided to three major steps: firstly, preprocessing process in which the original speech is transformed to digital forms by choosing proper sampling rate and record the Arabic speech words then separate this word to it's original Information (phonemes).Secondly, the global features of the Arabic speech phonemes are extracted using (Multiregdlet transform), this features are stores in data base that uses in testing and training to account for changes in phonemes. Finally, recognition of Arabic speech phoneme using Multi Layer Perceptron Neural Network (MLP), based on Feed Forward Back propagation used as classifier. The proposed system achieved a recognition rate within 99.4 %.
