The Research Of Speech Enhancement Based On Wavelet Transform And Neural Network

Speech signal is inevitably interfered by noise from the surrounding environment and transmission medium. The purpose of speech enhancement is extracted the pure original voice from the noisy speech.Wavelet transform theory is one of the modern analysis methods. In this paper, speech enhancement based on wavelet transform is systematically researched. the main idea of the wavelet threshold method is first, the noisy speech signal is transformed into wavelet domain with wavelet coefficients as output. Then a series of nonlinear mathematical treatment operate on the wavelet coefficients. Finally, the noise is removed by inverse wavelet transform, and speech enhancement is successfully realized.Wavelet algorithms on speech enhancement lie in several vital problems: threshold setting and the selection of threshold function. A perfect threshold function is also proposed to overcome the defects of traditional Soft-threshold function and Hard-threshold function. On the threshold setting, use the TEO Threshold estimates method. This method does not depend on the SNR estimation, noise frame detection, the implementation process is simple and effective. But at low SNR this method has the question of over the threshold. Combined the spectral subtraction and threshold-based adaptive algorithm TEO. First to use spectral subtraction to improve the signal's SNR, and then use TEO algorithm for speech enhancement, Avoid the problem of threshold over。In addition, use the wavelet function to replace the function of hidden layer neural network, to enhance the speech, in allusion to the complicated and large volume calculation of the structure of wavelet neural network using the difference evolution method to optimize network, increased the generalization of network.By using the Matlab software platform, we realize the algorithm, a large number of simulation results show that our proposed new threshold function method, spectral subtraction and TEO energy based on the method of combining algorithm for speech enhancement and Differential Evolution algorithm for wavelet neural network in both subjective and objective achieved good enhancement.