Exercise Activity Recognition with Surface Electromyography Sensor using Machine Learning Approach

Discriminating between various human activities has been the traditional focus of research on human activity recognition (HAR). This field of research is currently being applied to numerous daily activities, such as caring for the elderly, healthcare, smart homes, and athletics. A variety of types of sensors that can identify the different physical movements of humans are incorporated into most wearable smart devices, for example, smart armbands, smartphones, and smartwatches. Machine learning algorithms are employed by HAR in order to identify a range of activities in specific areas of normal everyday activities such as doing work and eating meals. Currently, exercise activity recognition (EAR) is regarded as an essential component in ambient assisted living, smart healthcare, and smart rehabilitation. Previous studies involving EAR have emphasized research on the improvement of the accuracy of the machine learning algorithms recognition that uses accelerometers and gyroscopes to provide sensor data. In this study, a framework of EAR using surface electromyography (sEMG) data is proposed. The results of the experiments indicate that the recognition accuracy can be improved by the use of sEMG data obtained from a combination of three sensors.