Research On Active Control Algorithm Of Impulse Noise

Noise pollution is considered to be the world recognized as one of the four major environmental pollution, it will be for people’s physical and mental health and survival environment have a huge impact.In recent decades, due to the traditional passive noise reduction (such as sound absorption, sound insulation, vibration isolation, etc.) in the low frequency noise reduction effect is not satisfied and the DSP technology fast development, the active noise control is considered to be a passive noise reduction in parallel with the traditional technology, and have very big concern and development.In the field of active noise control, the traditional linear adaptive controller with feedforward type filter-x least mean square algorithm in most cases, has obtained the good effect. But there are still many problems in practical use, one of them is when to eliminate noise for the impulse noise.Impulse noise with non-gaussian distribution characteristics, exists in many place in our lives, for example on the construction site of the pile driver, factory impact forging equipment, etc.This paper first analyzes the theoretical model, steady state distribution of impulse noise, introduces some existing pulse noise active control algorithm and analyzed their shortcomings.On the basis of the proposed filter-x nonlinear least mean square error (NCFxLMS) algorithm, the algorithm uses the compression of a nonlinear function transformation was carried out on the error signal, to convert the error signal to achieve the minimum of transient energy. And analyzes the convergence condition and the computational complexity of the algorithm. At the same time in order to overcome dead zone in the algorithm and error signal to control the compression degree effects, as well as the influence of the error signal to adjust the parameters of the compression of the nonlinear function in order to achieve faster convergence speed and better noise reduction effect.The algorithm to overcome the deficiency of the existing processing algorithm of impulse noise, do not need any priori information of impulse noise.Through theoretical analysis and simulation research, good convergence and stability of the algorithm.Then under the RLS structure and draw lessons from the thought of M estimator in robust statistics, proposed filter-x recursive least M estimator (FxlogRLS) algorithm, the algorithm by using M estimator is different and has 4 different algorithms.Given under the structure of RLS filter weight update very slow assuming no longer valid, on the basis of the original algorithm with algorithm is proposed to replace the sensor error measured error of real consider switching error filtering-x recursive least M estimator (FxlogRLSCE) algorithm.Then theoretically analyzed FxRLS, FxlogRLSCE, FxRLMECE algorithm 4 different types of seeds in the algorithm, and counted them in an iterative cycle of computation.Finally has carried on the simulation analysis of each algorithm, simulation results show that FXRLMECE algorithm is not only suitable for gaussian noise and impulse noise also is suitable for processing mixed noise, FXRLMECE algorithms in addition to the child algorithm based on Huber estimator, the rest of the algorithm are better than FxlogRLSCE algorithm.