Classification of various daily behaviors using deep learning and smart watch

In traditional healthcare and therapy, human behavior has been classified into only two categories: specific behavior and active behavior. As internet of things and wearable devices become popular, however, it is necessary to classify human behavior into more various categories for providing useful services. In this paper, we propose a novel classification scheme that classifies human behavior into 11 different categories including active and inactive activities in daily life. We collect data with smart watch and use deep learning model with a neural network for the classification. Extensive evaluation shows that various daily human behavior can be classified with 99.24% accuracy, and that the classification of human behavior can be used for various services.