Year: 2014
Type: article
Abstract: There is a growing demand on human activity recognition and remote wellness monitoring in modern societies.Related algorithms and hardware platforms have been intensively researched during the past decade.This work presents a wearable system for real-time human activity classification and fall detection.As the major part of this system, several algorithms are designed for feature extraction, vertical displacement estimation, and posture classification, and implemented in a customized wearable embedded platform.This battery-powered device has been tested on both undergraduate/graduate students and the elderly.The automatically logged activity reports in the testing are compared with video clips, manually logged activity, and the report from three different commercially available activity trackers.The testing results show up to 89.96% of second-by-second accuracy on daily activities, 75.2% fall detection sensitivity, and 99.85% fall detection specificity.Age Corr.Alg. 2 0.145 Height Corr.Alg. 1 -0.006Height Corr.Alg. 2 0.107 Weight Corr.Alg. 1 -0.079Weight Corr.Alg. 2 0.108
Author: Chia‐Hua Lin
Institution: Case Western Reserve University
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2724492330