心磁図における独立成分分析を用いた環境磁気ノイズ除去 : ―成分選択と縮約次元数決定の自動化―

概要

論文の詳細を見る
We have studied a method of environmental magnetic noise rejection using independent component analysis (ICA) for magnetocardiograms (MCGs). ICA is a useful method of separating the signal from noise. However ICA has various problems in determining numerous parameters and distinguishing signal or noise components from separated components. In many cases, these have been determined from the experiential results, and it has been difficult to determin the optimum parameters. We proposed a method of the distinguishing signal or noise components using the autocorrelation function of electrocardiograms (ECGs) and MCGs. Contraction dimension decisions were used to determin the maximum value of the autocorrelation function. The testing to verify its accuracy was carried out for MCGs with various signal-to-noise ratios. We found that this method was effective for the noise rejection.