Activity Recognition and Localization based on UWB Indoor Positioning System and Machine Learning

Joint activity recognition and localization plays an important role in many fields such as smart healthcare system, smart home, human-computer interaction, and robotics. Ultrawideband (UWB) is considered as a promising technology for high-precision indoor positioning system. But few studies have been done to simultaneously recognize and localize human activities based on the UWB indoor positioning system. In this paper, the possibility of simultaneously recognizing and localizing human activities with a self-developed UWB indoor positioning system is investigated. First, a few signal processing and machine learning techniques are applied to improve the positioning accuracy of the UWB indoor positioning system. Three machine learning methods based on support vector machine, artificial neural network, and hidden Markov model are then used to recognize five types of human activities based on the range measurements from the UWB indoor positioning system. Experimental results show that our approach achieves satisfactory performances in the joint activity recognition and localization task.