GA和RBF神经网络的Hilbert.Huang变换端点问题研究

Hilbert.Huang变换(HHT)在对信号进行经验模态分解和对各内禀模态函数进行Hilbert变换时会产生端点效应,端点效应会严重影响HHT的应用质量;为克服该问题,文中采用多目标分配遗传算法(GA)解决RBF神经网络(NN)模型训练的参数选择,并利用RBF_NN对信号延拓后再进行经验模态分解;该方法可有效克服经验模态分解方法的端点效应问题,得到具有明确物理意义的内禀模态函数和Hilbert谱;通过对典型确定信号和实际信号的仿真分析表明:文中提出的算法能有效解决HHT中存在的端点效应问题,且其效果优于RBF神经网络和支持向量机(SVM)的数据序列延拓方法。

Research of End Effects in Hilbert. Huang Transform Based on Genetic Algorithm and RBF Neural Network

The end effects of Hilbert-Huang transform are produced in the Empirical Mode Decomposition(EMD) and the Hilbert transform for Intrinsic Mode Functions(IMF),which have a badly effect on Hilbert-Huang transform.In order to overcome this problem,a multi-object allocation Genetic Algorithm(GA) to solve the parameters selection of RBF Neural Network(RBF_NN) is presented in this paper.Then the RBF_NN is used to predict the signal before EMD.The scheme can effectively resolve the end effects,and obtain the IMFs with explicitly physical sense and Hilbert spectrum.The simulation results from the typical definite and practical signals demonstrate that the end effects of Hilbert Huang transform could be resolved effectively,and its effects are better than prediction methods by RBF neural network and support Vector Machine(SVM),respectively.