Abstract
A new method of pulse signal de-noising based on wavelet transform and coherent averaging method is proposed. Pulse
signal is complex and weak, generally submerged by the interference of baseline drift, motion artifact and high
frequency noise. Consequently, it’s difficult to measure the heart rate by processing only one single-channel pulse signal,
especially when the noise frequency and the pulse signal frequency are in the same frequency range. In this paper, multichannel
pulse signal processing based on wavelet transform and coherent averaging is proposed to solve the above
problem. First, the detail coefficients and approximation coefficients of each channel signal are obtained by N layer
wavelet decomposition, then reconstructing the signal with high layers coefficients as the high frequency noises always
exist in low layers coefficients. In this way we can filter out the high frequency interference. Second, the centerline of
the upper and lower envelope curve obtained by cubic spline estimation is subtracted from each reconstructed signal so
as to eliminate the baseline drift completely. Finally, the heart rate is acquired with the coherent averaging method which
results in the noise being offset each other and the pulse signal being enhanced in the frequency range of pulse wave. The
pulse signal and three kinds of noise signals simulated with the superposition of different frequency sin signal were
analyzed, besides the experiment of six channel pulse signals measured simultaneously based on PhotoPlethysmoGraphy
(PPG) were conducted. The simulation and experiment results showed that this method was superior to the traditional
single channel.
