Extraction of human — object relations in Intelligent Space

This paper describes an observation system of human-objects interactions to extract relations among humans and physical objects in intelligent space (iSpace). The purpose is to obtain information of physical objects which can not be described in advance, in other words, depends on usage history of individual users. It is important issue for the iSpace to recognize human activities and design services for humans. In this paper, a method for multiple objects localization in the intelligent space is presented. The localization is realized by using two types of information. One is acceleration data which is used to determine whether an object is being operated. Acceleration data is acquired by using sensor network system. The other is hand position information of a nearest user to one object. By combining the information, positions of objects are able to be estimated by corresponding positions of user hands. In this method, determination of objects initial position is a first problem. In order to solve it, we use correlation of acceleration data among user hands and objects. Initial position of each object is given based on userpsilas hand position which is scored high correlation.