Font Size: a A A

Research On Signal De-noising Based On Wavelet Transform And Wavelet Neutral Networks Under Bimodal Noise

Posted on:2012-02-29Degree:MasterType:Thesis
Country:ChinaCandidate:R WangFull Text:PDF
GTID:2218330335485922Subject:Communication and Information System
Abstract/Summary:PDF Full Text Request
In recent years ,the study on non-Gaussian noise is one of the core contents of modern signal processing theory, and the signal processing theory of noise reduction are generally built on the basis of the Gaussian noise model, but in real-life we encounter with non-Gaussian signal. In this paper, non-stationary signal containing non-Gaussian noise is de-noised by the methods of modern signal processing, which has research and practical value.Starting from the characteristics of wavelet transform, we construct a new threshold function to de-noise signal with bimodal noise, the de-noised signal to noise ratio and mean square error indicators have improved considerably, so it is convenient to extraction of useful signal.In order to minimize the influence of the bimodal noise on the useful signal mode serious interference and improve signal to noise ratio (SNR), wavelet neural network is applied, wavelet neural network is a nonlinear filtering methods that can be used to reduce the noise adaptively. Improved gradient descent algorithm with morlet wavelet function has been used to train. In the processes of training, the neural network has a dynamic learning rate wavelet neural network, which can be adjusted automatically according to gradient descent. The experiment results show that neural network is effective methods in de-noising of signals. With gradient descent algorithm and morlet wavelet, the network has completed training in 59 times with the error 0.0167. Compared with normal gradient descent algorithm, the learning speed of the procedure here is evident increased. It avoids converging at a local minimum. The advantages show that wavelet neural network is a useful method in signal de-noise, it has a great significance in identifying useful signals and signal de-noising.
Keywords/Search Tags:Bimodal noise, Wavelet transforming, Wave neural network, Signal de-nosing, Non-Gaussian noise
PDF Full Text Request
Related items