Observing and Recording Human Activities

Computers play the role of environmental media, which supports various kinds of human activities. In order to support humans without disturbing their activities, computers need to recognize "situations" which are segments of the human activities, i.e. behaviors. In each situation, different support to human is required. Since the behaviors of human activities are observed by sensors, the definition and recognition of the situations have to be based on the sensor data. It is important to define categories of situations, observed image features useful to distinguish those categories, and the relationship between those categories and image features. In this article, we will describe the systems for observing human activities including lecture, e-learning and cooking as examples related to communication, knowledge acquisition and object manipulation respectively, because these three activities appear frequently in our daily life.