Realization of Long-Term Monitoring by a Home-Use Autonomous Mobile Robot Using Concept Drift Modeling

我们已经成功地构建了低成本的小型自主移动机器人，这些机器人可以在室内长期监控目标人，不包括更换电池等情况，并估计身体和心理的变化。有两种机器人。一种(4个机器人)由部件构成，另一种(2个机器人)基于机器人套件。每种机器人都配备了一台笔记本电脑和一台Kinect设备。我们已经开发了两种人体姿势监控应用程序和一个人脸监控应用程序。前者旨在防止摔倒，它会逐渐将大量观察到的骨骼数据聚集在一起，并发现相互相似的危险姿势。后者受心理疼痛检测的启发，通过聚类发现负面面孔。我们在相关研究方面也取得了成果。

We have succeeded in constructing low-cost, small autonomous mobile robots which monitor the target person indoor during a long term excluding occasions such as battery replacements and estimate physical and psychological changes. There are two kinds of robots. One kind (4 robots) was constructed from components and the other kind (2 robots) is based on a robot kit. Each robot of any kind is equipped with a notebook PC and a Kinect device. We have developed two kinds of human posture monitoring applications and a face monitoring application. The former, which aims at preventing falls, incrementally clusters a huge number of observed skeleton data and discovers dangerous postures which are mutually similar. The latter is motivated by psychological pain detection and detects negative faces which were discovered by clustering. We have also made achievements in relevant research.