Recognition Method of TCM Pulse-Conditions Based on Wavelet Packet Analysis and BP Neural Networks
基于小波包分析和BP神经网络的中医脉象识别方法*

Using the abalities of revealing the signal details and the local characteristics in the time-frequency domains, this paper presents a pulse-condition recognition method that is based on wavelet packets analysis and BP neural networks. The pulse-condition signals are decomposed into three layers wavelet coefficients by which the pulse-condition signals are reconstructed. On the third layer wavelet signals, the energy values of eight frequency bands from low frequency to high frequency are calculated. The energy values are used as the characteristic vectors of the pulse-condition signals, which are sent to improved BP neural networks as charateristic vectors to be trained. The experiment results of 480 pulse-conditions show that the recognition rate of our method is rather high.