Clustering approach to the problem of human activity recognition using motion data

Abstract

This paper describes authors' solution to the task set in AAIA'15 Data Mining Competition: Tagging Firefighter Activities at a Fire Scene (https://knowledgepit.fedcsis.org/contest/view.php?id=106). Method involves LDA classification on a preprocessed time series data with a unique label transformation technique using K-Means clustering. Data were collected from accelerometer and gyroscope readings.