Micro Activity Recognition of Mobile Phone Users Using Inbuilt Sensors

Human Activity Recognition using smartphone sensors is an area of active research. Micro activities of locomotion are indicators of higher level activities and general wellbeing of a user. In this paper, an approach for detecting a set of most common micro activities has been proposed and implemented. Five micro activities of locomotion namely sitting, standing, running, staircase ascend and descend have been considered. A two level classification model has been implemented to recognize these activities from data of inbuilt sensors of smartphone held in any of the three common positions by the user. Recognition accuracy of proposed approach is better than results reported in literature for similar problem. For purpose of training and testing, datasets have been collected on three different users and an android app has been developed to recognize activities in real time.