Extraction of Space-Human Activity Association for Design of Intelligent Environment

This paper presents a classification of human activities based on usage history of the spatial memory system and a numerical expression of each human activity in order to estimate individual human\´s intended purpose of an environment. We named the intended purposes for a place "space-human activity association". Intended purposes for a place will be different depending on the purpose of activities, even if observed activities are almost the same. Therefore, estimation of space-human activity association is important for intelligent environments to design services provided for individual human according to his/her current situation. The spatial memory system enables human users to store computerized information into the real world by assigning a three-dimensional position to the information, and to retrieve the information by directly indicating the point using their own hands. In the spatial memory system, what associates computerized information with a three-dimensional position is called "spatial-knowledge-tag (SKT)". The users actively create SKTs based on their need. As a result, arranged SKTs in a specified environment correspond to activity histories of the users. Also, users\´ intended purposes for the environment are reflected in the arrangement. Therefore, classification of arranged SKTs leads to classification of human activities, and accordingly estimation of intended purposes for places. In this paper, we describe two approaches to classify human activities. More specifically, the first approach is a classification of arranged SKTs by using only SKTs\´ information. The second approach is a classification of arranged SKTs by using both SKTs\´ information and usage history of the spatial memory system. As a result, the classification based on usage history of the spatial memory system showed multiple activities in the same area. On the other hand, the first approach resulted in a single cluster in the same area. We, then, give numerical values to SKT\´s c- - ontent type, and obtained numerical expression of each human activity