一种融合小波包分析和神经网络的信号消噪

The signal de-noising based on fusing wavelet packet analysis and neural network

为了改善非高斯噪声背景下信号的检测性能,将粒子群算法优化的神经网络(PSO-BP)和滤波器组重构的小波包变换(WPT)相融合,提出一种改进小波包神经网络(WPT-BP)的信号检测方法。首先选用Ackley函数验证PSO-BP具有较强的函数极值寻优能力,然后采用仿真信号序列检验改进WPT-BP检测信号的有效性,最后利用改进WPT-BP方法对MIT/BIH心律失常数据库中的心电(ECG)信号进行消噪。MATLAB仿真结果表明,改进WPT-BP在非高斯噪声且信噪比较低的情况下具有良好的消噪能力,是一种非常有效的信号检测方法。

In order to improve the detection performance of signals in the background of non-Gaussian noise, proposed the im- proved wavelet packet neural network（WPT-BP） of a signal detection method. This method fused particle swarm optimization and bank propagation of neural network （PSO-BP）which was optimized by particle swarm and filter bank reconstruction of Wavelet Packet Transform（WPT）. First the choice of the Ackley function validated PSO-BP with a strong ability to function extreme opti- mization. Then the validity of improved WPT-BP method was verified by the simulation of the signal sequence. Finally the MIT / BIH Arrhythmia Database of ECG signal was de-noising by using the improved WPT-BP method. The MATLAB simulation showed that the improved wavelet packet neural network had a good noise-canceling capability in the background of non-Gaussian noise and low signal to noise ratio. Thus it was a very effective method to detect signals.