Human fall detection with smartphones

According to the CDC (Centers for Disease Control and Prevention), one in three people over the age of 65 are likely to experience a fall. Twenty to thirty percent of these people sustain injuries such as fractures, loss of independence, and even death [1]. Fall detection is an active research area that strives to improve people's lives through the use of pervasive computing. This paper presents an approach to detect falls based on data gathered from a smartphone. It utilizes the smartphone's built-in sensors (accelerometer, gyroscope) to identify the location of the cellphone in the user's body (chest, pocket, holster, etc), and to find known patterns associated with falls. A general description on fall detection systems is provided, including the different types of sensors used nowadays. The proposed solution is presented and described in great detail. Finally, the system is assessed using known performance indicators. A total accuracy of 81.3% was calculated from the fall detection proposed algorithm. The top three locations to detect a fall were: texting with a 95.8% fall detection accuracy, pants' side pocket with an 87.5% accuracy, and shirt chest's pocket with an 83.3% accuracy.