Human periodic activity recognition based on functional features
Benyue Su, Jing Jiang, Qingfeng Tang, Min Sheng · 2016
In order to recognize the human daily activity more easily and accurately, an activity recognition method based on functional features is proposed in this article. Firstly, we transform the data series which collected by the wearable motion capture system into functional data using the techniques of Functional Data Analysis (FDA), thus the underlying continuity and periodicity of motion data can be depicted vividly. The method result indicates that functions of different activities indeed have different patterns. Secondly, a novel kind of feature called functional features including maximum, minimum and frequency of the function are put forward in order to classify different activities. Subsequently, SVM has been employed in order to achieve the highest accuracy according to the functional features. Finally, we compare our algorithm with other state-of-the-art methods so as to prove its effectiveness. Experimental results show that our algorithm can describe the continuity and periodicity of the motion capture data precisely and recognize human periodic activity accurately.