Physical activity classification using body-worn inertial sensors in a multi-sensor setup
- 14 November 2016
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 10, 1-4
Abstract
Physical inactivity significantly impacts personal health, reduces quality of life, and often leads to mobility disorders, diabetes, and cardiovascular disease. Monitoring daily life activities by means of wearable inertial sensors can provide valuable feedback necessary to improve the quality of daily life and prevent the development of mobility disorders caused by physical inactivity. In this study, a physical activity classification (PAC) algorithm was developed and tested using an inertial sensor-based dataset. The dataset was acquired from multiple inertial sensors, each mounted at a different body location, and consists of various Activities of Daily Living (ADL). Data from nineteen healthy young subjects were analyzed. Time- and frequency-domain features from raw 3D accelerometer and 3D gyroscope signals were computed by performing windowing of the time series data. The K-nearest neighbors (KNN) pattern recognition algorithm was used to classify thirteen different ADLs and was evaluated by a 10-fold cross-validation. The proposed PAC algorithm outperformed the existing algorithm validated using the same dataset, with an overall mean classification rate (sensitivity) of 97.38%. This paper discusses the limitations of this study and proposes ways to overcome said limitations in order to make the PAC algorithm more effective in real-life conditions.Keywords
This publication has 17 references indexed in Scilit:
- Instrumented shoes for activity classification in the elderlyGait & Posture, 2016
- Human Activity Recognition Based on the Hierarchical Feature Selection and Classification FrameworkJournal of Electrical and Computer Engineering, 2015
- A random forest classifier for the prediction of energy expenditure and type of physical activity from wrist and hip accelerometersPhysiological Measurement, 2014
- Activity recognition with smartphone supportMedical Engineering & Physics, 2014
- Evaluation of accelerometer based multi-sensor versus single-sensor activity recognition systemsMedical Engineering & Physics, 2014
- Computational Methods for Estimating Energy Expenditure in Human Physical ActivitiesMedicine & Science in Sports & Exercise, 2012
- SHIMMER™ – A Wireless Sensor Platform for Noninvasive Biomedical ResearchIEEE Sensors Journal, 2010
- A Triaxial Accelerometer-Based Physical-Activity Recognition via Augmented-Signal Features and a Hierarchical RecognizerIEEE Transactions on Information Technology in Biomedicine, 2010
- Preprocessing techniques for context recognition from accelerometer dataPersonal and Ubiquitous Computing, 2010
- The WEKA data mining softwareACM SIGKDD Explorations Newsletter, 2009