Design Of Speech Enhancement Algorithm Based On Wavelet Transform

Information interaction technology has been rapidly developing,and peoples' needs and requirements for information interaction equipment have been constantly increasing.Nowadays,the communications between people and people,people and machines,people and equipment,sending instructions,operations,etc.are inseparable from the voice.Since the development of human society,the voice has always been an important medium for communicating information to each other.Speech is a type of random signal,which is featured by sometimes variable and non-stationary.Therefore,in daily communication between peoples,the voice signal is very susceptible to interference from the surrounding environment,causing the receiver to be unable to accurately receive information.Not only that,all kinds of interference signals generated by the external environment will cause the inaccuracy of the received voice information during the transmission.In this context,this paper uses the idea of wavelet analysis to study the noisy speech signal.The detailed works are as follows:1.Through the studies on the characteristics and principles of the two basic signal processing algorithms,as well as the learning of various previous speech enhancement algorithms,wavelet transform algorithms are combined with the wavelet transform thinking to process speech signals.Among them,the wavelet threshold denoising algorithm is the main analysis object of this article.Also,the wavelet threshold denoising algorithm is comprehensively analyzed.2.Combining the advantages and disadvantages of traditional algorithms and the algorithms of later scholars,an improved wavelet threshold function is proposed.In the traditional threshold function,the hard threshold function algorithm has natural flaws in the establishment of the algorithm,which will cause oscillations after the signal is processed.Although the signal can be effectively avoided after being processed by the soft threshold function,the voice signal is seriously distorted.In this paper,the new algorithm adds new adjustable parameters to optimize the threshold selection.In the face of different decomposition levels,the threshold value is adjusted to adapt to different decomposition levels.On the basis of solving the problems of traditional threshold function,the denoising effect of the algorithm is further enhanced.3.In this paper,a new denoising method combining Kalman filter with improved algorithm is proposed.The new joint de-noising algorithm can maintain high-level denoising ability,being a feasible enhancement algorithm.