Classification of Various Daily Activities using Convolution Neural Network and Smartwatch

Abstract

In the traditional human activity recognition field, human activity has been classified into two categories: exercise type and exercise posture. However, as the internet of things technology and wearable devices have been developed and become popular, in order to provide useful services, it is necessary to classify daily activities as well as existing activities. In this paper, we propose a novel classification model that classifies human activities into 11 different categories including activities that are highly active and less active in daily life. We collect data with an off-the-shelf smartwatch and use a deep learning model with a convolution neural network for the classification. An extensive evaluation shows that various daily human activities can be classified with 97.19% accuracy.