Research And Application Of Signal De-Noising Via Wavelet Analysis

Signal and information processing is one of the most rapidly developing disciplines of the Information Science in the recent twenty years. Signal processing includes: signal de-noising, feature extraction, edge extraction. Signal denoising is most common in the signal processing. Limitations have their own application of classical signal denoising method such as the pure time domain method, the pure frequency domain method, Fourier transform, Fourier transform window. The wavelet transform is developed in the nineteen eighties a new time-frequency analysis method, it has good localization characteristic in both time domain and frequency domain, and have been widely used in the signal to noise in wavelet transform.There are many de-noising methods of wavelet transform, and threshold-shrinking is used most widely. However, the classical threshold shrinkage method has its own defects, or the reconstruction signal have constant deviation with the original signal, or the presence of reconstruction signal will appear additional concussion. Another important aspect of threshold de-noising is the selection of threshold. If the threshold is too big, it will misjudgment a part of the useful signal as noise and filter it out, that will make reconstruction signal loss some information; If the threshold is too small, it will keep part of the noise wavelet coefficient, so the de-noising is not complete.In addition, once you start de-nosing based on the "first generation wavelet" de-noising method, you can only choose a wavelet, and the de-noising process can’t accord to the partial feature of the signal to change the wavelet. However, different wavelet has its own characteristic and applicable signal, therefore, if we can be choose a wavelet from a wavelet set according to the local features of the signal, to reach a local optimum and ultimately achieve the overall best. It is a very challenging and at the same time, it also represents the development direction of the wavelet de-noising topics.This paper tries to analyze and research above three problems, and proposes a new threshold function; this threshold function has good flexibility and better de-noising effect. In addition, the threshold selection scheme has been improved, so that the threshold can be dynamically selected based on the wavelet decomposition level. Finally, on the basis of the theory of ascension, this paper puts forward an adaptive method, this method can dynamically select wavelet according to the partial feature of the signal, and make use of each wavelet’s advantages, achieve a good de-noising effect.