Home // AMBIENT 2012, The Second International Conference on Ambient Computing, Applications, Services and Technologies // View article

Hierarchical Human Activity Recognition Using GMM

Authors:
Prateek Srivastava
Wai-Choong Wong

Keywords: Activity recognition; wearable sensors; pattern recognition; Gaussian Mixture modeling (GMM).

Abstract:
Abstract— In this paper, we propose a hierarchical human activity recognition system using Gaussian mixture models (GMMs) on continuous daily activities. The system recognizes the human activities by making use of tri-axial accelerometer and bi-axial gyroscope. We use different features such as mean, variance, root mean square, pitch, and roll for activity classification. Comparative performance assessments are carried out using the publicly available Wearable Action Recognition Dataset (WARD). The hierarchical recognition happens in two steps. First, the test data is classified into two broad clusters – static activity and dynamic activity. Second, the recognition is carried out within the identified class. For continuous activity recognition, our proposed system is able to achieve a recognition accuracy of 86.92% which is 2.63% above the baseline system. The new algorithm also provides more flexibility for better feature selection for different sets of activities.

Pages: 32 to 37

Copyright: Copyright (c) IARIA, 2012

Publication date: September 23, 2012

Published in: conference

ISSN: 2326-9324

ISBN: 978-1-61208-235-6

Location: Barcelona, Spain

Dates: from September 23, 2012 to September 28, 2012