인자분석에 의한 생체신호 신호대잡음비 향상법 및 기존의 방법과의 비교

Abstract

Electronic signals from biological systems such as electrocardiogram(ECG) sometimes are obscured by random noises. When the signal-to-noise ratio(S/N) is low significantly by certain amount of noises, it is sometimes very difficult to get correct biological information from patients. Conventionally, polynomial smoothing methods such as Savitzky-Golay filters are employed, but the efficiency of S/N enhancement is limited. Usually, enhancement factor of S/N is small, at most with the best set of parameters for the polynomial smoothing. As a new solution for the noise reduction, intrinsic ability of factor analysis is applied for the noise reduction, which is separation of meaningful eigenvectors from those of noises. In this study, clean ECGs were generated and certain levels of root-mean-squares(RMS) noises were mixed. These noisy ECGs were subject to the conventional method of noise reduction and a new method of factor analysis and the results of S/N enhancement were compared. As a result, the best enhancement of S/N calculated from levels of RMS noises was achieved by factor analytical method.