A Novel Motion And Noise Artifacts Reduction Mechanism (MNARM) For Wearable PPG-based Heart Rate Extraction

Heart rate which means the number of heart beats per minute is a very important health indicator in many fields.It can help with health care or movement monitoring.But until now,most heart rate monitors have many limitations,such as the subject must be stationary or at a fixed location,which means the real-time monitoring of heart rate anywhere is not easy.Photoplethysmography(PPG)is a noninvasive measurement of the change of the vessel volume at the surface of the skin.Oxygen saturation and heart rate can be calculated by analyzing the amplitude,rhythm,in periodicity of the PPG signals.The size of PPG sensor is usually small,making it suitable to be embedded in the wearable device,which makes the real-time monitoring of heart rate possible.However,there are still many challenges to get the heart rate through the PPG signals processing.As be mentioned above,PPG signals are biological signals extracted from surface of skin,so it is hard to extract the signals due to low intensity.On the other hand,high-frequency,electrical noise,and motion noise could cause a great impact on the quality of the collected signal.Motion artifacts are caused by tissue effect and venous blood changes during body movement,they are the most difficult to eliminate due to their major frequency components could overlap with those of heart rate.There are some researches that trying to solve the problems.But there are still some defects in this area.In the aspect of algorithm,Independent components analysis(ICA),wavelet,and empirical mode decomposition method were only suitable for offline processing due to their high complexity and demand for huge data source.For online processing,different kinds of adaptive filters are widely used.But usually an off-line synthetic noise reference signal are taken as the compensation which would cause a great amount of computation for wearable PPG device.In the aspect is hardware,accelerometers were employed to track motion,and usually PPG signals were collected from the fingers.Such devices were too big to wear comfortable.In this paper,a novel MNARM is presented to remove motion artifacts and noise artifacts of PPG signals and calculate heart rate in real time based on a wireless,wrist wearable PPG device.The device we designed includes a 6-axis accelerometer to improve the accuracy of the acceleration signal acquisition.MNARM combines NLMS(Normalized Least Mean.Square)adaptive filtering algorithm and Mallat algorithm,removing artifacts of different frequencies respectively and effectively.We designed three experiments to show performance of MNARM at effect of removing artifacts,individual difference and exercise intensity.The results were as follows:the MNARM was very efficient,and greatly weakened the motion and noise artifacts as can be seen from the comparison of signals of different stages;the device was adaptable to different subjects in which error rate was controlled at less than 5%;the device could also achieve 5%margin of error,when the velocity of motion was less than 8.5km/h.