Human Activity Recognition System using Smart Phone based Accelerometer and Machine Learning

摘要: Human Activity Recognition (HAR) has gained significance importance due to its wide range of applications in security, healthcare, surveillance, virtual reality, control systems and automation. Sensors embedded modern mobile phones enable unobtrusive detection Activities Daily Living (ADL). Various statistical deep learning techniques for the automated human activity have been presented recently. In this study, we collected accelerometry data through a phone carried by user number days classify ADL on basis exhibited movement into stationary, light ambulatory, intense ambulatory abnormal classes. such as walking, sitting jogging etc. are performed classified simultaneously application users comparative analysis. Collected is given an input trained model analyzed implementing J48 classifier. Results reveal accuracy score around 70% each class it noted that classification was with above 80% stationary activity. It shown can be recognized high using constrained environment single sensor. classifier also correctly activities strong correlation between them chair standing position. This work significant utilization long term health monitoring capable ensuring neurological masses HAR accelerometers.