Physical Activity Classification with Smartphone based Accelerometer, Gyroscope and Device Motion for Personal Diabetes Healthcare Management

Abstract

Physical inactivity has a substantial negative influence on one’s health, lowers quality of life, and frequently causes cardiovascular disease, diabetes, and mobility problems. Both diabetes and the patient’s lifestyle have a significant impact on each other. Although we shouldn’t overburden most diabetes patients with technology since they can manage their condition without it, lifestyle-monitoring technology can nevertheless be helpful for both patients and their doctors. As a result, we created a method of lifestyle monitoring that makes use of smartphones, which the majority of patients already have. In this study, we demonstrate our smartphone-based system that uses the accelerometer, gyroscope and GPS incorporated into smartphones as sensors to identify, categorise, and rate running, walking, laying and standing activities. On two publicly accessible data sets, namely the UCI HAR data set and the motion sense data set for a physical activity sensor, several classification analysis approaches are explored. In comparison to previous efforts, our classification model technique significantly improves the classification of various activities. The proposed gated recurrent unit (GRU) architecture have an average accuracy of 94.91% in classifying activities.