Spoken Arabic Digits Recognition Using Discrete Wavelet

In this paper, we propose a scheme for recognizing isolated spoken Arabic digits, based on the Discrete Wavelet Transform (DWT) features. The Discrete Wavelet Transform is a transformation that can be used to analyze the temporal and spectral properties of non-stationary signals like audio, based on the time-frequency multi-resolution property of wavelet transform. In this paper, the extracted wavelet coefficients from audio speech used in the recognition of spoken Arabic Digits. The performance of the Discrete Wavelet Transform as recognition features for recognizing spoken Arabic digits is considered for different scales of wavelet signal and different Wavelet families. In general, the recognition of spoken Arabic digits with Daubechies wavelet family is more efficient than all other tested wavelet families. In our experiments we note that the recognition performance of spoken Arabic digits using wavelet family is not very effective.