Adaptive de-noising in arterial pulse wave based on lifting scheme discrete wavelet

Abstract

Pulse wave denoising is an essential procedure in pulse wave analysis. Lifting wavelet denoising speeds up the typical wavelet denoising, and is thus of great interest. Three groups (five each) of data sets (radial pulse waves recorded from healthy subjects and sphygmogram and plethysmogram obtained from the Multi-parameter Intelligent Monitoring in Intensive Care database) were enrolled for this study. The predictor of the lifting scheme were adaptively calculated using the Least Mean Square (LMS) algorithm. Comparison analysis were applied with the typical wavelet denoising and adaptive denoising using typical wavelet. The adaptive denoising algorithm using lifting scheme can effectively eliminate the noise in pulse wave signal (MSE is 0.0469, 0.0256, 0.0147, 0.0088, 0.0051 and 0.0035, respectively when the SNR of the pulse signal equals 5, 10, ···, 30db). As the SNR gets higher, the performance of the adaptive denoising algorithm using lifting scheme gets closer to those of the typical wavelet denoising and adaptive denoising algorithm using typical wavelet (MSE of the lifting scheme denoising algorithm and the other two typical algorithms, 0.0035, 0.0036 and 0.0084, respectively, with SNR of the raw pulse signal 30db).