A Millimeter-Wave MIMO Radar Network for Human Activity Recognition and Fall Detection

Falling is a major risk for elderly people. To enable independent living, fall detection and activity monitoring are desirable. Radar is a sensor principle that offers the possibility to detect falls in a contactless, privacy-preserving fashion. Therefore, in combination with deep learning, it has become a widely investigated technique for human activity recognition and fall detection. Current systems, however, come with some limitations: When using just one monostatic radar, it is impossible to measure lateral velocities. This motivates the use of a radar network consisting of two spatially orthogonal radars. Contrary to some previous works which applied similar radar networks, this paper introduces the first millimeter-wave multiple-input-multiple-output (MIMO) radar network with two orthogonal radars for human activity recognition and fall detection. Using millimeter-wave MIMO radars enables a higher resolution and the use of angular information for the recognition task. First measurement results and deep-learning-based activity recognition are presented.