Human Activity and Posture Classification Using Wearable Accelerometer
This paper presents an analysis of several machine learning classification models and their effectiveness in classifying human activity and posture using data gathered from wearable accelerometers. This analysis was done with a publicly available dataset comprising 1656,633 samples organized into five activity classes, provided by researchers at the Pontifical Catholic University of Rio De Janeiro. In particular, this study focuses on the efficacy of Logistic Regression Classification, Stochastic Gradient Descent Classification, Random Forest Classification, Support Vector Machines and Artificial Neural Networks.
Wunderlich, K. & Abdelfattah, E. (2018). Human activity and posture classification using wearable accelerometer data. 9th IEEE Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON), pp. 77-81, doi: 10.1109/UEMCON.2018.8796831.