摘要
为了实现中医脉象的客观、准确分类,文章提出了一种基于BP神经网络的脉象识别方法。考察了隐层节点数对网络收敛速度、识别正确率的影响以及学习率对收敛速度的影响,改进了网络训练算法。并选取了较好的学习率参数对脉象信号进行了网络训练,获得了满意的网络收敛误差和识别精度。最后用大量临床脉象样本对网络和算法进行了检验,实验结果表明该方法能够实现对中医常见脉象的准确、快速分类。
In order to achieve the objective and exact classification of the traditional Chinese medicine pulse-conditions,a pulse-condition recognition method based on BP neural network has been successfully developed.The influence to convergent speed and recognition accuracy of both concealed layer node count and the learning rate is considered.The BP network-training algorithm is improved,which obtains satisfactory network convergent error and recognition effect.The recognition tests of many pulse-condition samples indicate that our method works well in classify the traditional Chinese medicine pulse-conditions veraciously and quickly.
出处
《计算机工程与应用》
CSCD
北大核心
2005年第32期187-189,共3页
Computer Engineering and Applications
基金
国家自然科学基金项目(编号:70471057)
陕西省自然科学基金项目(编号:2001X28)
关键词
中医脉象
模式分类
神经网络
BP算法
pulse-condition,pattern classification,Neural Networks,Back Propagation algorithm