Human activity recognition based on random forests

Abstract

Human activity recognition is a hot topic in the field of pervasive computing and context aware computing, and may support lots of potential applications, such as healthcare, smart home, etc. In this paper, a novel human activity recognition method is proposed, where accelerometer data are collected from wearable devices and used as input of random forests (RF) model to classify human activity. Random forests model for human activity recognition is established, and the algorithm is designed and analyzed as well. Experiment results proved our method valid, which achieved an overall accuracy about 90%.