Physical Activity Recognition Based on Rotated Acceleration Data Using Quaternion in Sedentary Behavior : A Preliminary Study

Shin, Young Eun ;   Choi, Woo-hyuk ;   Shin, Taemin

This paper suggests a physical activity assessment method based on quaternion. To reduce user inconvenience, we measured the activity using a mobile device which is not put on fixed position. Recognized results were verified with various machine learning algorithms, such as neural network (multilayer perceptron), decision tree (J48), SVM (support vector machine) and naive bayes classifier. All algorithms have shown over 97% accuracy including decision tree (J48), which recognized the activity with 98.35% accuracy. As a result, physical activity assessment method based on rotated acceleration using quaternion can classify sedentary behavior with more accuracy without considering devices' position and orientation.