COMPARATIVE ANALYSIS OF MACHINE LEARNING ALGORITHMS FOR OPTIMIZING VARIABLE STEP-SIZE LEAST MEAN SQUARE IN MOTION ARTIFACT REDUCTION

Optical sensor like Photoplethysmographs (PPG), is widely used in generating real time information such as current heart rate. Existing studies on PPG demonstrated that the weakness of this technology is the sensor will capture the motion artifact reading when there is excessive motion exerts on the sensor. Numerous algorithms had been developed to reduce the motion artifact on PPG and increase the accuracy of the health monitoring device reading. However, these existing solutions using least mean square (LMS) algorithm failed to achieve high accuracy of heart rate reading. This paper presents and compares three types of machine learning algorithms that are widely used in classification of wearable signals, which are support vector machine (SVM), artificial neural network (ANN) and random forest (RF). The machine learning algorithms optimize variable step-size LMS (VSSLMS) accuracy by classifying the speed of the motion and giving suitable step size values based on the classification. The result shows that SVM is the best machine learning algorithm in classifying the speed category of the heart rate to eventually get the suitable step size value for VSSLMS.