A Feature Set for the Similar Activity Recognition Using Smartphone

Abstract

Human activity recognition(HAR) has been a hot topic in recent years. With the development of smartphone and sensor technique, using smartphone to recognize human activity seems possible. However, due to the diversity of activities, it is not easy to distinguish them, especially for the similar activities. In this paper, we focus on the HAR with the smartphone. We construct a 5-element vector according to the outputs from the 3-axis accelerometer to eliminate the effect of orientation. Then we study the features of activity data from three aspects time domain, frequency domain and time-frequency domain. After that a feature set is chosen for the similarity activity recognition. Experimental results show that this feature set works well under the Multi-Layer Perception classification, even for the general daily activities.