Classifying motion using radio waves and machine learning
LE3 .A278 2014
2014
Zhang, Haiyi Silver, Danny
Acadia University
Master of Science
Masters
Computer Science
Over the past three decades it has become increasingly apparent that the global population is aging. A larger elderly population and fewer medical professionals have placed an increasing burden on professionals and families. The most common injury su ered by the elderly is falls. Numerous solutions that allow a fall victim to call for help are available, but current solutions require some interaction by the user to operate and care for the devices. This work investigates the use of machine learning to analyze radio waves to determine if patterns emerge in the signal that can be used to di erentiate between fast and slow motions, and then using the same principles to detect when a person falls or lays down. We present a system that consists of two antennas to transmit and receive radio waves, software to extract relevant data from the waves, and a machine learning system to classify motions that requires no interaction by the elder. Our solution is capable of classifying falls with a 90% accuracy when classifying the falls of one person, and 73% accuracy when using multiple participants. We conclude that received signal strength alone cannot be used to create an accurate model, but other characteristics of radio waves potentially o er better results and are left for future investigations.
The author retains copyright in this thesis. Any substantial copying or any other actions that exceed fair dealing or other exceptions in the Copyright Act require the permission of the author.
https://scholar.acadiau.ca/islandora/object/theses:315