超高齢社会を迎える我が国において，独居高齢者の安心・安全な生活の維持は重要な課題である．地域の医療・支援体制が整っていない過疎地域では，独居高齢者の日常的な見守りは行政コストの観点から重要といえる．既存の高齢者見守りの研究ではウェアラブルデバイスやカメラを使用しているが，対象者にデバイスの装着を強いることや，動画として監視することによるプライバシの侵害や心理的ストレスなど課題があげられる．本研究では対象者のプライバシに配慮した人感センサを用いた見守りシステムを提案し，対象者ごとの行動パターンに合わせた見守りを行うため，対象者ごとに確率ネットワークモデルを構築し，異常検知を行う．独居高齢者の家屋に人感センサモジュールを設置し実証実験を行い，人感センサモジュールを設置した部屋間での遷移確率を指標とすることで，医療従事者の知見をふまえて異常検知を行った．実験結果より，独居高齢者の日常生活に現れる多くの症状を推察することができ，本システムの優位性を示した．

In Japan, a super-aging society, maintaining the safety and security of the elderly living alone is an important issue. In depopulated areas where local medical care and support systems are not in place, daily monitoring of the elderly living alone is important from the perspective of administrative costs. Existing studies on monitoring of the elderly have used wearable devices and cameras, but they have issues such as forcing the subjects to wear the devices and invasion of privacy and psychological stress by monitoring them as videos. In this study, we propose a monitoring system using a human-detecting sensor that considers the privacy of the subject, and constructs a probabilistic network model for each subject to detect anomalies in order to perform monitoring according to the behavioral patterns of each subject. We conducted a demonstration experiment by installing a motion sensor module in the house of elderly people living alone, and detected anomalies based on the knowledge of medical professionals by using the transition probability between rooms where the motion sensor module was installed as an indicator. The experimental results showed that the system was able to estimate many symptoms that appear in the daily lives of elderly people living alone, indicating the superiority of the system.