基于小波分析和BP神经网络识别的中医脉象信号辨识系统

讨论基于小波分析的脉象信号分解与重构、信号除噪、脉象信号时频特征值的提取和分析，构建合理的神经网络结构，各层神经元数量的确定，选择合理的学习速率，脉象信号特征值的选取，神经网络的训练等必须解决的关键问题。构建了一种比较实用的基于小波分析BP神经网络的中医脉象信号辨识系统。经1456例临床脉象检测，准确率〉90％，不仅极大地提高对平、浮、沉、迟、数、虚、实、滑、涩、洪、弦、促、结、代等基本脉的识别率，对于由上述基本脉构成的临床常见的相兼脉也有相当高的识别能力。

System for Recognition of Chinese Medical Pulse Signal Based on Wavelet Analysis and Back Propagation Neural Network Recognition

Based on wavelet analysis, this paper discusses the cruxes which must be solved , including pulse signal decomposition, reconstruction and signal purification as well as extracting and analyzing pulse eigenvalues of time domain and frequency domain, and how to construct reasonable neural network configuration, how to make sure the numbers of nerve cells at all levels, how to select suitable study speed and efficiency ,how to train neural network and so on, and constructs a considerably practical system for recognition of Chinese medical pulse signal based on wavelet analysis. The accurate rate of pulse detection of 1456 clinic cases is proved high by 90%, even higher. It significantly enhances not only the recognition rates to basic pulse signals such as normal pulse, floating pulse, sunken pulse, slow pulse, fast pulse, weak pulse, powerful pulse, slippery pulse, astringent pulse, full pulse, rapid and intermittent pulse, slow and intermittent pulse, slow -intermittent-regular pulse, but also the recognition capacity of clinically common combined pulse signal with basic pulse.