改进样本熵最优小波包阈值选择算法在信号降噪中的应用

Improved Optimal Threshold Selection Algorithm Applied to Denoising

在小波包样本熵理论基础上,结合改进粒子群优化算法选择最优阈值,以样本熵为判据,对小波包的各个分解层设置不同的阈值,构造阈值步长随时间变化的函数,选取噪声序列样本熵最大的阈值作为最优阈值。对某引信发射过载信号进行分析,证明了该方法选取小波包阈值的可行性与有效性。未改进算法之前存在较大的人工选择阈值步长的误差,改进优化算法之后最优阈值误差在允许范围内,而且降噪效果更好。

An improved particle swarm optimization algorithm was proposed to select the optimal threshold based on the wavelet packet sample entropy theory. Taking the sample entropy as the criterion, the method set different thresholds for each decomposition layer of the wavelet packet. It constructed the function of the threshold step with the change of time, and the maximum entropy threshold was selected as the optimal threshold. The feasibility and effectiveness of the method of selecting threshold was proved by analyzing the overload signal of a certain fuze. There is a large error of artificial selection threshold step before the improved algorithm. After improving the optimization algorithm, the optimal threshold error is within the allowable range, and the effect of denoising is better.