Nonlinear filters for preprocessing Heart rate variability signals

摘要

Heart rate variability analysis requires normal sinus rhythm to accurately acquire heart rate variability measures in the time and frequency domain. Ectopic beats, missed QRS complexes and noisy beats hinder this analysis and introduce ambiguity in the variability measures. For this reason, it is necessary to design a specific filter to remove the ectopic beats and other noisy beats from HRV signal. In this paper a nonlinear adaptive threshold based Rank Order Filter (AROF) is proposed for denoising HRV signal. The filter has adaptability in rank, window size and threshold conditions based on the noise level. The quality of the restored signal is measured by the standard time and frequency domain measures of HRV signal and general statistical measures such as peak signal to noise ratio (PSNR) and root mean square error (RMSE) of restored signal. The performance of adaptive threshold based AROF is compared with wavelet based filter, median filter and an adaptive median filter. The performance of adaptive threshold based Rank Order Filter is superior not only in the PSNR value but also in the quality of the restored signal.