Research On Adaptive Noise Cancellation System Based On Neural Network

With the development of modern society,the problem of noise elimination has attracted wide attention.How to effectively eliminate noise has become an urgent problem.Usually,LMS adaptive filter and RLS adaptive filter are used to deal with the noise.However,it's not ideal for nonlinear problems while the ability of nonlinear approximation of neural network can solve these problems well.Therefore,it is very important to study the adaptive noise cancellation system based on neural network.Firstly,the principle of adaptive filter and the principle of adaptive noise cancellation technology are introduced in the paper,and the performance of adaptive noise cancellation system is analyzed.Then,the LMS algorithm and RLS algorithm are introduced,and it is pointed out that the neural network algorithm has obvious advantages through a brief comparison between them.Secondly,the principle of BP network and the structure of RBF network are introduced.Several BP algorithms,RBF learning algorithms and the parameters of BP network are emphatically analyzed.The influence of BP network algorithm,parameters and RBF network parameters on the performance of adaptive noise cancellation system is simulated by experiments.Under the condition of linear correlation and nonlinear correlation of two noise signals,a BP network adaptive noise cancellation system based on OSS algorithm and a RBF network adaptive noise cancellation system based on OLS algorithm are established.The specific research contents include determining the input signal,BP network algorithm,the number of hidden nodes in the BP network,the transfer function,the maximum node number of hidden layer of RBF,and the diffusion speed of RBF.And then,under the condition of linear correlation and nonlinear correlation of two noise signals,the BP network adaptive noise cancellation system based on OSS algorithm and the RBF network adaptive noise cancellation system based on OLS algorithm are respectively simulated with different input signals.Then the time domain charts and the error variation curves of the signals after noise cancelling of the two systems are analyzed,and the differences of system performance are compared.The output SNR and SNR increments of the two systems after noise cancellation are calculated,and the SNR increment is compared.The experimental results show that the two systems have good noise cancellation effects in both linear and nonlinear correlation problems.The RBF network adaptive noise cancellation system based on OLS algorithm has better performance in dealing with the nonlinear correlation problem of two paths noise,and the BP network adaptive noise cancellation system based on OSS algorithm performs better under the condition of linear correlation of two noise signals.Finally,the RBF network adaptive noise cancellation system based on OLS algorithm is used to extract fetal ECG signals.Simulation results show that it can cancel interference noise in real time and extract fetal electrocardiogram signals with good results.