Kinect-based fall detection system for the elderly

Kong, Tian Chyuan (2024) Kinect-based fall detection system for the elderly. Project Report. Universiti Teknikal Malaysia Melaka, Melaka, Malaysia. (Submitted)

Text (Full Text) Kinect-based fall detection system for the elderly.pdf - Submitted Version Download (2MB)

Abstract

The growing population of elderly individuals are increasing in correlation to the advancement in healthcare. Along with that, however, are the associated risks, such as incidents of falls. Several studies had shown that older adults experienced at least one fall every year, and it was the main cause of accidental death in older adults aged 65 or more. In Malaysia, many elderly are left alone during the day as their family members are out to work or school. Therefore, a fall detection system was designed to detect when a person experiences a fall or a loss of balance. The system utilizes Kinect sensors and algorithms to detect the movements and postures of an individual, aiming to analyze and identify patterns that indicate a fall event. Skeletons and joints such as heads, shoulder centre, hip centre, ankle left, and right are detected and extracted. The fall algorithm is implemented to obtain the y-coordinate values and threshold values. Several observed fall scenarios include falling to the left side, falling to the right side, falling to the front, falling to the back, and falling while sitting. The result of lower accuracy at short distances from the Kinect sensor can be attributed to its limited field of view and depth perception issues at close range, leading to incomplete or distorted skeleton tracking. Slightly longer distances provide a more optimal range for accurate skeleton tracking and fall detection, while accuracy increases at longer ranges due to increased detail in the captured data. The fall detection accuracy of 100% is obtained throughout the evaluation of all heights of the Kinect sensor and lighting conditions.

Actions (login required)

View Item