Adaptive Algorithm for Motion Artifacts Removal in Wearable Biomedical Sensors During Physical Exercise

Motion artifacts (MAs) are significant sources of noise in wearable devices that are used to collect biological information from the human body. Photoplethysmography (PPG) is an example of the sensors embedded in many of these devices. PPG signals are highly influenced by MA, especially when the motion involves the organ to which the PPG sensor is attached. The aim of this article is to build an adaptive algorithm and multiresolution analysis (MRA) technique to denoise the raw PPG signals, suppress the MAs and accurately estimate the heart rate (HR) as an example of MA suppression in wearable devices. The challenge here is to find the right selection considering high accuracy and low-computational complexity. Discrete wavelet transform recursive inverse (DWT-RI) adaptive filter algorithm in addition to MRA is implemented to suppress MAs and estimate the HR using the simultaneously recorded accelerations (ACCs) in the three dimensions as reference signals for the adaptive algorithm. The proposed algorithm has a low computational cost and results in an absolute average error of 1.17 beats/min when tested on 12 subjects from a well-known database for this purpose. The performance of the proposed algorithm is compared to the other existing methods in the literature used for noise removal and MA suppression. DWT-RI adaptive algorithm can successfully suppress MAs in PPG signals with low computational complexity and outperforms several alternatives in terms of the average absolute error.