Abstract: PPG signal is an effective method to assess the cardiovascular parameters like heart rate, blood oxygen saturation, blood pressure and respiration rate. Motion and other artifacts introduced during acquisition of PPG, limits the accuracy in estimation of clinical parameters. In this paper, we present wavelet based feature extraction and motion artifacts removal methodology to overcome the issues of heart and respiratory parameter estimation. In the proposed algorithm, point of interest coefficients are ascertained for every wavelet sub-band grid, producing a changed wavelet sub-band network. This makes the exhibited algorithm more efficient. Every information set comprises of deliberately made conceivable MA noises, viz., vertical, flat, waving, and squeezing MAs with distinctive breathing examples. The technique is connected on the recordings accessible of Physionet dataset. The measurable and classification investigation, performed to test the viability of the introduced new algorithm, uncovered a decent acknowledgment for inferred respiratory signal, when contrasted and the initially recorded respiratory signals utilizing traditional strategy. To perform the classification of the dataset, neural network is used. The proposed strategy obviously outperforms the customary heart rate detection system in the vicinity of MA. The results of proposed system show the approximate 99.9% of sensitivity and 99.7% of specificity.

Keywords: PPG, Heart rate, Respiration rate, classification, neural network.