Recognizing User Activity Using a Smartphone's Accelerometer and Deep Neural Network Classifier

Along with today's fast-growing technology, machines/devices especially mobile devices have been developed using many sensors to simplify the user's activities. One of the most known and frequently used sensors is called accelerometer, daily used as a step counter, image stabilization, and user interfaces control. However, activity recognition is considered a difficult task due to the reality that each activity has its unique features and there is no clear analytical way to analyze sensor data into specific forms of action in general. This study examines the potential and exciting ability of the accelerometer to recognize user activity by making simple prototype to support the implementation of this user activity recognition. After data acquisition, deep learning classifier will be used to differentiate activities. This research will show the efficiency and utilization of using accelerometer combined with deep learning in recognizing user activity, which can be associated with many applications for advance study such as falling detection, abnormality detection, and prediction of human behavior.