The Research Of Speech Signal Enhancement Method Based On Wavelet Transform

There exist various disturbs from surrounding environment, noises caused by transform medias, inner electrical noises of the communication equipments, even disturbs from other speakers during the speech communication process. Speech enhancement is to improve the speech signal quality, and improve the articulation and intelligibilityIn practical application processing, the signals are no steady, especially for the speech signal section which are strict no steady, the detail features were blurred by using traditional methods. The wavelet transform which have the character of time-frequency localization and the nature of the multiple resolution ratios analysis provide an effective tool for above problems. The wavelet transform can decompose speech signal by the multiple scales. For the character of the speech signals and random noises under different scales, so it' s good to the denoising for speech signals.In this thesis, speech signals enhancement methods are studied base on wavelet transformation.(1) The thesis mainly introduces methods for speech signals enhancement and analyses their advantage and weakness, envisages the basic theories of wavelet transformation and application prospect in wavelet denoising.(2) The wavelet denoising basis principles and methods using wavelet are analysised and studied, the advantages and disadvantages of these methods are summarized.(3) Mainly studies threshold function in the wavelet threshold denoising. In the wavelet threshold function, to solve the insufficiency of the soft and hard threshold, a novel threshold function is proposed, and the analysis and derivation are presented.(4) Specially modifies the select of threshold, which are important for the wavelet denoising. With Neyman-Pearson Criterion and character of speech signals, a novel wavelet speech denoising method based on the energy cell and new threshold rule is proposed in this thesis.