Real-time human activity recognition in smart home with binary tree SVM

In this paper, binary tree support vector machines (BT-SVM) and a sliding sensor window have been applied to recognize the human activities in real time. To construct the optimal binary tree, genetic algorithm (GA) has been employed. The real data collected from participants performing activities was used to train the SVM models and evaluate the accuracy of the recognition algorithms. Compared with one-against-one SVM and BT-SVM with human knowledge, the experiment results show that the proposed BT-SVM with GA has better activity recognition performances in smart home.