KiFall: Privacy Preserving Fall Detection using Smartphone and Wireless Camera Sensors
The occurrence of a gait imbalance or a fall while walking is a large health issue for at-risk populations, such as seniors, individuals convalescing from brain strokes, among others. Fast and precise detection of a fall is hence of critical importance. Current existing solutions consist on wearable sensors placed at different body locations, which can be expensive, intrusive and uncomfortable. In this work we propose an indoor, fast, privacy-preserving fall monitoring framework using off-the-shelf components such as smartphones and wireless cameras, implemented with an Android-based smartphone and the Microsoft Kinect. Our approach follows a two-step system, allowing an early trigger on the smartphone, which then activates the camera. If a fall is detected, the system sends out an an emergency alert. Many experiments have been developed to decide which sensors on the smartphone best capture the abnormal data traces associated with a fall event, whether to process the raw data, the impact on the processing time, etc. Experimental results and comparison with other popular fall detection applications reveal improved performance (false alarms and miss detections below 4%), fast response time (7 seconds) and limited impact on battery life (15h before consuming the battery) through our dual-stage approach.
Keywords: body area networks, mHealth, computational sensing, statistical data analysis, framework development, client-server architecture
Technologies: Android, Kinect, Matlab, TCP Server, wereable sensors
Programming languages: Java, C#
Technologies: Android, Kinect, Matlab, TCP Server, wereable sensors
Programming languages: Java, C#
----
Check out: Thesis
Paper for IEEE Transactions on Information Technologies in Biomedicine under revision
Source codes on github: Android app, Kinect app, TCP server
Check out: Thesis
Paper for IEEE Transactions on Information Technologies in Biomedicine under revision
Source codes on github: Android app, Kinect app, TCP server
MS Thesis in Computer Engineering, Supervised by Prof. Kaushik Chowdhury, Mathworks SMART Laboratory, Northeastern University, Boston, USA, Oct 2012 - Sept 2013.