Activity Recognition in New Smart Home Environments

Abstract

Activity recognition is important to health care in smart homes. It provides information about the activities of the residents. Many health care services are based on it. To collect data about the activities, sensor networks consist of binary sensors are widely used. Activity recognition is performed based on their readings. Most of existing activity recognition methods are based on supervised classification algorithms. One drawback of these methods is that the classification model learned in one smart home environment usually cannot be used in another. For a new smart home environment, sufficient sensor readings have to be collected and labeled to learn the needed classification model. This process is time consuming and expensive. In this paper, we propose a method for smart home activity recognition with binary sensors. Our method utilizes the characteristics of binary sensors, the semantic information of the sensors and the activities, and the time information. The classification model learned with the data in one smart home environment can be used in the activity recognition in another, which has different sensor networks and label spaces. Experiments on real world datasets show the effectiveness of our method.