Fall risk assessment and visualization through gait analysis

Abstract

Falls are a major health concern among elderly populations. There is a critical need to develop automated systems for assessing a patient's fall risk although the methodologies for determining this risk vary in efficacy, accessibility, and comfort. With advancements in smart home technol-ogy, aging in place and accurate fall risk assessment are no longer mutually exclusive. This paper presents a user friendly fall risk assessment system designed for care providers to non-invasively but continuously monitor their patient's risk of falling. The proposed system employs a pressure sensor-embedded floor - a SmartFloor - installed in the patient's home to monitor trends in gait pa-rameters like gait speed, stride length, and step width. The system should allow care providers to visualize dangerous changes to their patient's gait 24/7 and without disturbing the patient. How-ever, falls are few and far between, making it hard to evaluate how effective fall prediction systems are. To facilitate diagnoses and fall risk assessment, the system also reconstructs a skeletal visu-alization of each recorded walking segment. This is done using a motion similarity algorithm and a database of SmartFloor and Microsoft Kinect data. We tested the accuracy of several variations of the motion similarity algorithm using a small pool of seven participants and the results are presented in this paper.