Body activity recognition using wearable sensors
Long Cheng, Chenyu You, Yani Guan, Yiyi Yu · 2017 Computing Conference · 2017
Efficient recognition of human body activities is of great significance in many fields. With the development of wearable sensor technology, wearable sensors are playing a more and more important role in recognizing human body activities. How to accurately and quickly recognize human body activities while saving computing resource is a major challenge to be addressed. Utilizing data sampled from wearable sensors, this paper solves the human body activity recognition problem from perspectives of both machine learning and compressed sensing. Specifically, three different machine learning algorithms, artificial neural network, support vector machine and hidden Markov model, and one compressed sensing involved algorithm, sparse representation classification method based on random projections, are used to recognize human body activities, respectively. Meanwhile, various numerical experiments based on a real-world dataset collected using wearable sensors are conducted to validate the effectiveness of these algorithms. Numerical results demonstrate that all of the four algorithms achieve satisfactory recognition performance and the sparse representation classification method based on random projections outperforms the other three machine learning algorithms.