Human activities recognition using smart phones

Abstract

Embedded sensors in smartphones provide real-time information of users' movements and activities. This becomes base for human activity recognition to made smart environment where user can monitor their physical health and life style properly. Many prominent machine learning techniques are available that processes this sensor data to derive the high-level information of users, such as their locations, activities that they are performing (e.g. walking, running, stationary, walking downstairs, walking upstairs etc.) and the associated energy devices they are using (e.g. computer, coffee machine, washing machine etc.). Through an integration of these sensor data from mobile devices, this project aims to develop an integrative information system concerning the analysis of simple and complex activities the occupants are performing within a smart living environment and their usage of energy related equipment. Machine learning techniques, Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), K-Nearest Neighbor (KNN) and Random Forest (RF), are used in this work.