Research Of Multi-mode Noise Suppression Based On Adaptive Filter

Any system are subject to varying degrees of noise interference, noise reduction and suppression of research was a hot topic in the field of modern communication. Currently many research for gaussian noise assumption, but in practice we encounter a non-gaussian noise, gaussian noise by non-linear system, its system of output but also for non gaussian noise. The result will make system performance degradation, and serious damage and loss of signal, unable to extract the desired signal. This paper using adaptive noise suppression system and research on new algorithms of elimination and suppression for the noise.This paper first describes the basic concepts of multi-modal noise and its mathematical model. Then to explains the fundamentals of adaptive filtering, and lists of common adaptive filtering algorithm, such as least mean square algorithm (LMS algorithm) and some improved LMS algorithm. This article uses the adaptive noise reduction system, it is a typical application of adaptive filtering. In multi-mode background noise, some algorithms do not apply, so this new algorithm study on suppression for the noise. In the case of less noise pollution, combined with the advantages of the Griffith's LMS (GLMS) algorithm and LMS-Newton algorithms make the GLMS-Newton algorithm and improved. This algorithm not only with variable step size based on cross-correlation factor, and introduces is based on input signals and gradient operators expect cross-correlation of the signal; In cases with low signal-to-noise ratio(signal seriously contaminated by noise), present a suppression for the noise of a mixed-norm filtering algorithm (NMNF), cost function introduces a hybrid in this algorithm parameters, By adjusting the mixing the best filtering algorithm, and the establishment of step-size parameter with the nonlinear relationship of changes of the input signal-to-noise ratio.Paper finally adopted to propose new algorithms for the simulation of MATLAB simulation software, comparative analysis of the stability and convergence of the algorithm performance, and so on.