中医脉冲信号特征盲提取算法设计

Design of blind extraction algorithm for pulse signal characteristics of traditional Chinese medicine

针对中医脉冲信号特征在混叠噪声信号中提取时存在盲源分离性能较差、均方误差较大,收敛性能不佳的问题,提出并设计了基于人工蜂群算法的中医脉冲信号特征盲提取算法,该算法对传感器阵列捕获的所有中医脉冲观测信号做中心化处理和白化处理,利用互信息和信息熵的关系,计算得到预处理后中医脉冲信号特征提取代价函数,并将串音误差作为中医脉冲信号特征盲分离性能度量指标;在此基础上,按照人工蜂群算法中引领蜂搜索-跟随蜂搜索-侦察蜂搜索步骤求解中医脉冲信号特征提取代价函数以获得最佳分离向量,实现中医脉冲信号特征盲提取。仿真对比结果表明,提出算法克服了当前算法盲源分离性能较差、均方误差较大,收敛性能不佳的问题。

At present,when the feature of TCMS pulse signal is extracted from the aliased noise signal,the performance of blind source separation is poor,and the mean square error is large,and the convergence performance is not good.Therefore,this paper presented and designed an algorithm of blind extraction for TCMS pulse signal feature based on artificial bee colony algorithm.This algorithm centralized and whitened all TCMS pulse observation signals captured by the sensor array.Based on the relationship between mutual information and information entropy,we obtained the cost function of feature extraction of TCMS pulse signal after the pretreatment.And then,we used crosstalk error as the blind separation performance measures of TCMS pulse signal feature.According to steps in the artificial bee colony algorithm such as leader search-follower search-scout search,we found the cost function of TCMS pulse signal feature extraction,so as to obtain the best separation vector.Finally,we achieved the blind extraction for TCMS pulse signal features.Following conclusion can be drawn from simulation results.The proposed algorithm overcomes the problem of poor performance of blind source separation,large mean square error and poor convergence.