Pulse wave digital processing and beat detection based on wavelet transform and adaptive thresholding

This paper considers the adaptive algorithm for pulse wave processing in the presence of various physiological interferences, such as motion artefacts and baseline wander. The proposed processing technique consists of two main steps: comprehensive filtering of pulse wave based on wavelet transforms and noise-resistant pulse beats detection based on the set of pass-band filtering, nonlinear transforms and adaptive thresholding. To eliminate baseline wander from pulse wave signal we introduced a new algorithm based on the principles of adaptive noise cancellation, while the reference signal for adaptive filtering was generated by using multiresolution wavelet transform. For reducing motion artefacts and others high frequency distortions of pulse wave we suggested an approach based on soft thresholding of detailed coefficients from wavelet decomposition. The efficiency of the developed algorithm for pulse wave signal processing was assessed in comparison with the existing approaches by using mathematical modeling of pulse wave signal and occurred contaminations. The performance of the pulse beats detector was further verified for different recordings of clinical pulse wave signals from the Physionet database.