Human fall detection using a circular audio sensor array

Abstract

[ACCESS RESTRICTED TO THE UNIVERSITY OF MISSOURI AT REQUEST OF AUTHOR.] According to the U.S. Census Bureau estimate (March 2010), there were 38.9 million people 65 years and older in the United States in 2008. In 2005, 15,800 people 65 years and older died, and an estimated 1.8 million people 65 years and older were treated in emergency departments for nonfatal injuries from falls. In order to offer help in a timely manner following a fall, many researchers and scientists have already proposed several human fall detection devices. The challenging task is to design a fall detector that is robust, accurate, not intrusive and generally accepted by old adults. In this work, we propose a circular microphone sensor array for detecting falls and generating a message to caregiver and medical supporter automatically. This novel fall detection system leverages the existing research and expertise from the Center for Eldercare and Rehabilitation Technologies (CERT) at the University of Missouri. This approach does not require users to manually activate or wear a sensor device. In previous work, members of CERT developed a linear microphone array. In this work, we are interested in a circular audio sensor array that has improved fall location. We have a stunt actor perform fall for us and we test our system by conducting a set of comparison of experiments. The results of our experiments confirm that the fall detection performance of our system is promising.