用于信号特征提取、信号消噪，通过反复训练模板能有较高的识别率，基于分段非线性权重值的Pso算法，matlab开发工具箱中的支持向量机，预报误差法参数辨识-松弛的思想，验证可用，采用加权网络中节点强度和权重都是幂率分布的模型，详细画出了时域和频域的相关图。- For feature extraction, signal de-noising, Through repeated training zISCHRClate have higher recognition rate, Based on piecewise nonlinear weight value Pso algorithm, matlab development toolbox support vector machine, Prediction Error Method for Parameter Identification- the idea of relaxation, Verification is available, Using weighted model nodes in the network strength and weight are power law distribution, Correlation diagram shown in detail the time domain and frequency domain.