Experiment-Supported Mobile Application for Monitoring Human Activities Using Neural Networks

Abstract

During the last years, people have been interested to understand the human nature by creating smart algorithms and systems which can detect or recognize activities performed by humans. In this paper, we propose the implementation of a smart system able to recognize and monitor human’s activities. The system implemented as an application running on a smartphone using Android OS. The mobile application determines what activity is performed by a person and stores the information into an internal database. The database can be used to generate statistics. By using statistics, the users can check the records of their daily activities. Our mobile application can detect 4 types of activities: standing, sitting, walking and jogging. For determining these activities, we used AI software implemented in PyTorch. The AI inputs consist in coordinates read from the smartphone’s accelerometer. To evaluate the performance, we tested the neural network in 2 different modes: using a testing set and when it is part of the mobile application (someone performs an activity and the application detects it).