Activity/Posture Recognition using Wearable Sensors Placed on Different Body Locations

Hristijan Gjoreski and Matjaž Gams


Artificial intelligence, intelligent systems, data mining, ambient intelligence, activity recognition, posture recognition


This paper presents an approach to activity/posture recognition using wearable sensors. The sensors consist of 3-axis accelerometer, 3D gyroscope and 3D magnetometer (3D compass). Several approaches of activity and posture recognition are analyzed, varying the number, type of the sensors and using different body placements. The effect of these different scenarios on the accuracy is investigated. Placements of up to three sensors are considered: chest, right thigh and right ankle. Extraction of additional attributes was performed in order to more precisely reconstruct the activities/postures. The results indicate that the gyroscope and magnetometer data significantly improve performance of the system, in particular when only one sensor was placed on the chest.

Important Links:

Go Back