脉象特征人工神经网络分类器

Human Pulse Features Processing System Based on Artificial Neural Network

以人工神经网络为手段，以提取脉象信息为目的，由临床采样数据形成了网络训练输入特征向量库，不以单一脉本身为分类对象，而考虑它是否是某些可识别特征的组合，建立了浮沉、弦滑、迟数等一组脉象特征网络。证实了人工神经网络用于具有模糊性的脉象特征的识别和分类是可行的，带智能处理的特色。其分辨准确率可达９０％。

An intelligent analysis system which can distinguish insignificant difference in pulse was developed to apply artificial neural network to adapt the fuzzy feature in the Traditional Chinese Medicine pulse analysis. 50 clinical samples were collected to train and test the system. The neural networks were trained to recognize a group of features in the different pulses instead of pulses themselves. Sample database was set up, pulse features was extracted and pulse feature classification neural networks were formed. The diagnosis accuracy is up to 90%. Thus, it is feasible to use neural network in pulse processing.