Purpose -Real-time multimedia supervisory systems generally include a distributed system that delivers live video input captured with cameras at the watched person's site, using a PC or hand-held device at the distant supervisor's site. The system comprises many entities such as cameras for image capture, transmission software, network connections, receiver software, a display device, multimedia processing software and hardware, control software, etc. The purpose of this paper is to realize a safe and convenient supervisory system that autonomously provides users with services that fulfill users' requirements related to quality and privacy in a ubiquitous information environment. Design/methodology/approach -A system is designed by integrating environmental information acquired from the real world and knowledge related to human social activities. A real space understanding mechanism is proposed to infer the situations and relationships of users by combining sensing information and social knowledge. Social knowledge related to human relationships, the life style of the watched person, home structure, etc. is used with ontology to infer the situations of users. Findings -An early prototype was implemented for supervising elderly people and some experiments were performed based on several scenarios. Results obtained from some experiments confirmed that this supervisory system can provide real-time multimedia supervisory services for elderly people, with reasonable quality-of-service and privacy that meet the users' requirements. Originality/value -The system described in this paper assesses the situation of users and surrounding environmental information to provide appropriate supervisory services. This paper provides insight into the design and development of ubiquitous application systems to realize comfortable and safe services using a combination of environmental information and social knowledge.