Combining adaptive filtering and motion state for wearable PPG-based heart rate extraction

Photoplethysmography (PPG) technology is a method to detect the change of blood volume in subcutaneous tissue, which is characterized by simple principle, low cost, non-invasive and sustainable monitoring, and is widely used in heart rate monitoring of wearable devices. However, the original signal obtained by the single-channel PPG sensor contains a lot of noise, especially the motion artifacts generated in the process of motion. In this paper, the acceleration data and synthetic signals in different directions are used for adaptive filtering, and the heart rate data is modified by combining motion state estimation. This method improves the robustness of adaptive filtering. We conducted a test in which ten individuals simultaneously wore watches with three-axis accelerometers and single-channel PPG sensors, Polar heart rate armbands (Verity Sense), and Polar heart rate chestbands (H10) while performing various physical activities such as jogging, running, and jumping for 30 minutes. The three-axis acceleration signal is utilized as the reference signal for adaptive filtering and estimated the user’s motion state, and the heart rate change rate and range are determined according to the motion state, which can correct the heart rate result. Compared with Polar heart rate chestband data, the average absolute error of the entire sample was 7.5 beat per minute(BPM).When the amplitude of motion is not particularly intense, the algorithm proposed in this paper can identify heart rate more precisely.