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 p