INVESTIGATING THE GAIT PATTERNS IN FALLING-DOWN PREDICTION FOR THE ELDERLY USING MACHINE LEARNING
BRENDA KIM
Provert Research Laboratory, 111 Charlotte Place Suite#100, Englewood Cliffs, NJ 07632, USA.
INSEO ANGELA CHOI
Provert Research Laboratory, 111 Charlotte Place Suite#100, Englewood Cliffs, NJ 07632, USA.
RYAN PARK
Provert Research Laboratory, 111 Charlotte Place Suite#100, Englewood Cliffs, NJ 07632, USA.
JONGBIN LEE
Provert Research Laboratory, 111 Charlotte Place Suite#100, Englewood Cliffs, NJ 07632, USA.
TAEHYUK KIM
Provert Research Laboratory, 111 Charlotte Place Suite#100, Englewood Cliffs, NJ 07632, USA.
*Author to whom correspondence should be addressed.
Abstract
Many falls usually lead to chronic complications for the elderly. Four common causes of falls include slippery floors, the level of brightness, stairs, and residential obstacles such as a rug. These are linked to household activities, and therefore, the fatal falls often occur when an elderly person stays home alone, a timely contact to the doctor is not possible.
Objective: The study aimed to apply a dynamic functionality embedded in a microcontroller to detect true falls and activate alarms, promptly.
Methods: The accelerometer embedded in Arduino NANO 33 IoT measured the 3-axis acceleration in the gait cycle. The 3-axis acceleration characterized the dominant frequency and mean peak. These two characteristics could distinguish between real falls and fake falls. Actual falls were defined as the ability to continue moving after a fall. Acceleration data was then analyzed using the double integration to find the foot clearance in the four most common causes of falls.
Results: The study demonstrated that foot clearance was decreased in the four situations and that Arduino NANO 33 IoT could accurately distinguish between real falls and fake falls, proving the device's ability to detect falls that constitute an emergency.
Conclusions: The potential for the Arduino Nano 33 IoT was confirmed to detect falls in senior citizens through this study. The Arduino Nano could distinguish between real falling and fake falling, regardless of whether it is walking or running before the fall.
Keywords: Accelerometer, cost function, fast Fourier transform, gait pattern, household activities, minimum foot clearance, motion tracking, senior falls.