Detection of Falls and Activities of Daily Living from Inertial Measurement Unit Sensors Worn by Subjects with Parkinson’s in a Free-Living Environment

Student: Maxwell Sakyi

Project Mentors: Dr. Narayanan Krishnamurthi – CONHI
Dr. James Abbas- SBHSE
Dr. Claire Honeycutt- SBHSE

YouTube Link: View the video link below before joining the zoom meeting

Zoom Link: https://us04web.zoom.us/j/7727498281

Zoom meeting time: 10am - noon

Abstract

Parkinson’s disease (PD) is a progressive neurodegenerative disorder that affects movements of daily life and takes a toll on independent living. One specific health risk factor for individuals with PD is the increased likelihood of falls resulting in major injuries (40%-60%), minor injuries (30% - 40%) and fractures (5% - 6%) ; up to 20% of these falls lead to death subsequent to hip fracture. The use of wearable sensors such as Inertial Measurement Units (IMUs) enables synchronized data collection from tri-axial accelerometers, gyroscopes and magnetometers which can provide information on the quality of the movement during activities of daily living. This study is directed at automatic detection of falls, walking, turning, standing and sitting events in a free-living environment. Data sets were collected from the daily activities of subjects diagnosed with PD with the IMUs placed over each foot and at the center of the lower back. Falls were detected by identifying the data that satisfy three different thresholds: a resultant angular velocity of 14 rad/s, the resultant change in trunk angle of 1.7 rad/s2 and a resultant angular acceleration of 0.08 rads/s2. Walking segments were detected by a 0.025 rad2/s2 threshold of the maximum spectral coherence magnitude obtained from the conjugates of the Fourier transform of gyroscopes on the left and right feet between 1.5Hz to 2.5 Hz. Turns were also detected using appropriate thresholds on the yaw axis, while sitting and standing were detected with the pitch axis, all from the lumbar gyroscope. The automatic identification of these events during activities of daily living would help to develop falls prediction techniques and rehabilitation techniques to avoid/correct movements that can increase the risk of falls.