Noise Filtering Using FFT, Bayesian Model and Trend Model for Time Series Data

Mitsue ONODERAa, Yoshimi ISUa, Umpei NAGASHIMAa,c*, Hiroaki YOSHIDAa, Haruo HOSOYAa and Yuuzou NAGAKAWAb,d

aFaculty of Science, Ochanomizu University,
Bunkyo-ku, Tokyo, 112-8610 JAPAN
bHelth Care Center, Ochanomizu University,
Bunkyo-ku, Tokyo, 112-8610 JAPAN
cPresent address: National Institute of Materials and Chemical Research, MITI
1-1 Azuma, Tsukuba 305-8565, JAPAN
dPresent address: Mitsubishi Electric Co. Ltd

(Received: December 4, 1997 ; Accepted for publication: March 3, 1998 ; Published on Web: June 21, 1999)

The applicability of noise filtering methods, FFT, Bayesian model and trend model was evaluated using multiple nonstationary frequencies of noise and electrocardiogram with large signal-to-noise ratio. The results show that the Bayesian model is the most applicable to filtering of multiple nonstationary frequencies with large noise. In the case the best way for filtering out the noise in electrocardiogram suggested that the back ground rolling noise should be eliminated by FFT after removing the high frequency component by Bayesian or trend model.

Keywords: Noise filtering, FFT, Bayesian model, Trend model, Time series

Abstract in Japanese

Text in Japanese

PDF file(555KB)

Author Index
Keyword Index