Research On Speech Enhancement Algorithm And Implementation Of MATLAB

Speech is interfered with various kinds of noise and the voice of the other speakersin the actual process of obtaining and transmitting it. These interferences may affect thequality of speech signal, or even affect the realization and performance of the otherspeech signal processing systems. Speech enhancement which is an effective method toimprove the speech quality is to reduce the noise in speech signal, and it has becomeone of the core technologies in processing speech signal.The research of this paper is based on the elementary knowledge and the existingalgorithm of speech enhancement. The main research contents and innovations areshown below:Firstly, the realization of the speech enhancement method based on the modulusmaxima algorithm and threshold value algorithm of wavelet transform is on the basis ofthe wavelet denoising principles. The results of simulation experiment indicate that thede-noising results of the two methods are all good. The modulus maxima algorithm andthe threshold value algorithm are all good for removing noise when the signal mixedwith white noise; but the de-noising results of modulus maxima algorithm is better thanthat of the threshold value algorithm when there appears more singularity in the signal. Secondly, the improved LMS algorithm is on the basis of the principle of adaptivefilter, the steepest descent method and the traditional LMS algorithm. The results ofsimulation experiment indicate that the improved LMS algorithm can remove the noiseefficiently when they take different steps. But there will be some difference in the effectof filter and convergence rate.Finally, the realization of the adaptive noise canceller is based on combining thewavelet denoising with adaptive filter technologies, which makes the adaptive noisecanceller can be used in case of having no noise reference source. In this method, thenoise components should be isolated through denoising noisy signals with waveletdenoising method at first, and then eliminate noise in the adaptive noise canceller ofwhich the input is the isolated noise components. The results of simulation experimentindicate that the improved adaptive noise canceller can eliminate noise efficiently, and itis more convenient to apply it to practice.