Abstract:
Physical activity recognition has been an important issue over the past decades, and has become a very important technology in a wide range of applied research. Using triaxial accelerometer sensors to collect huge amounts of motion and posture data during a relatively long period of daily life in subjects’ personal living environment has been proved to reliably detect basic human motions (DeVaul and Dunn, 2001), and be accurate enough to determine postures of the human body (Foerster et al., 1999). To date, researchers around the world, especially those from the area of biomedical engineering, have made great progresses in recognising human motions, including physical activities of daily living and falls in the real world, by using accelerometer sensors. Plenty of algorithms have been invented for research purposes (Bagalà et al., 2012, Karantonis et al., 2006) and some of them have been implemented and integrated along with accelerometer sensors for real-time human movement monitoring (Karantonis et al., 2006). However, to evaluate the accuracy of activity recognition algorithms during research or before integrating them into real sensors, there are still some extra and redundant work that people have to do, which are out of the scope of their core research goals. For example, to manage the experimental data, to visualise signals, to validate the recognised activities against the original signals, to adjust the parameters of algorithms, and to analyse the trend of human activities and so on. These kinds of tasks are usually time consuming and may annoy researchers and professionals, especially those without much IT knowledge. This thesis presents the design and implementation of an extensible software system, Activity Recognition Engine or ARE, which addresses the issues above. It takes advantages of a number of technologies, including data visualisation, database, web 2.0, signal processing, multi-threading, and multi-processing. The implemented system is platform independent and based on Browser-Server architecture. So, technically, everyone is able to use the system without installing any additional software on their PC desktops or laptops, as long as they have a generic web browser and are connected to the network.