超定独立分量分析及其在结肠压力信号分析中的应用

Over-determined ICA and Its Application on the Analysis of Human Colonic Pressure Signals

针对传感器个数大于源信号个数的情况,提出了一个超定独立分量分析(ICA)方法,并将其应用到人体结肠压力信号分析中.首先对结肠压力信号运用主成分分析(PCA)估计源信号个数,并将观测信号进行降维使其维数与源信号相等,从而将超定ICA问题转化为一个完备ICA问题;最后通过快速ICA算法估计出源信号.实验采用26组人体结肠压力数据,成功提取出结肠活动中具有生理意义的动力模式信号.

Independent component analysis（ICA） is a broadly used method for blind source separation（BSS）.The mature ICA model has a restriction that the number of sources must equal that of sensors used to collect data,which is hard to meet in practice.An over-determined ICA method was proposed and successfully used in the analysis of human colonic pressure signals.Using principal component analysis（PCA）,the method estimated the number of sources and reduced the dimension of observed signals so that it is the same with that of the sources;and then,FastICA was used to estimate all the sources.From 26 groups of colonic pressure recordings,colonic motility patterns were extracted,which not only proved the effectiveness of this method,but also would greatly facilitate further medical researches.