多模噪声背景下基于遗传小波网络的滤噪研究

Research of De-Noising Based on Ga—Wnn under Multi-Mode Noise

为了改善小波神经网络(WNN)存在收敛速度慢、易陷入局部极小值的缺陷,将WNN和遗传算法(GA)相结合,提出一种基于遗传小波网络(GA—WNN)的多模噪声中确定信号的滤噪方法。由于该方法融合了WNN良好的时频局部分析能力和GA自适应全局快速寻优的特点,因此,GA—WNN不仅克服了WNN存在的不足,而且可以使WNN参数最优化,从而进一步提高WNN的滤噪性能。仿真表明,在多模噪声背景下,GA—WNN能有效地从含噪信号中提取确定信号,并比传统的WNN滤噪效果好。

In order to improve wavelet neural network（WNN）deficiency,which is in a slow convergence rate and easy convergence to local minimums,and it has combined genetic algorithm and wavelet neural network to form GA-WNN,which are used to de-noising the multi-mode noise.Because this method has the good local performance in both time and frequency fields of the WNN and the adaptive fast global searching ability of the GA that GA-WNN not only overcomes the shortcoming of the WNN,but also can make the WNN parameter optimal.Therefore this method can further improve the filtering performance of the WNN.Simulation results show that GA-WNN can effectively extract the determine signal from the noisy signal and is better than the traditional WNN de-noising effect in the context of multi-mode noise.