Physical Activity Recognition from Smartphone Embedded Sensors
The ubiquity of smartphones has motivated efforts to use the embedded sensors to detect various aspects of user context to transparently provide personalized and contextualized services to the user. One relevant piece of context is the physical activity of the smartphone user. In this paper, we propose a novel set of features for distinguishing five physical activities using only sensors embedded in the smartphone. Specifically, we introduce features that are normalized using the orientation sensor such that horizontal and vertical movements are explicitly computed. We evaluate a neural network classifier in experiments in the wild with multiple users and hardware, we achieve accuracies above 90% for a single user and phone, and above 65% for multiple users, which is higher that similar works on the same set of activities, demonstrating the potential of our approach.
KeywordsPhysical activity accelerometer classifier smartphone
Unable to display preview. Download preview PDF.
- 2.Boyle, M., Klausner, A., Starobinski, D., Trachtenberg, A., Wu, H.: Poster: gait-based smartphone user identification. In: Proceedings of the 9th International Conference on Mobile Systems, Applications, and Services, MobiSys 2011, p. 395. ACM Press (June 2011)Google Scholar
- 4.Martin, J.J.: Recognition of motion patterns based on mobile sensor data. Msc, University of Stuttgart (2010)Google Scholar
- 8.Longstaff, B., Reddy, S., Estrin, D.: Improving activity classification for health applications on mobile devices using active and semi-supervised learning. In: Proceedings of the 4th International ICST Conference on Pervasive Computing Technologies for Healthcare, pp. 1–7. IEEE (2010)Google Scholar
- 10.Weiss, G.M., Lockhart, J.W.: Identifying user traits by mining smart phone accelerometer data. In: Proceedings of the Fifth International Workshop on Knowledge Discovery from Sensor Data, SensorKDD 2011, pp. 61–69. ACM Press (August 2011)Google Scholar