Human body recognition based on the sparse point cloud data from MIMO millimeter-wave radar for smart home

Human body recognition is widely used in smart home. The current mainstream perception modalities, i.e., camera and wearable device, are vulnerable under challenging lighting conditions and poor convenience. On the other hand, Multi-human body recognition remains as one of the most challenging tasks in a dynamic and complex environment. In this work, we introduce the low-cost multiple-input-multiple-output (MIMO) millimeter-wave radar without exposing user's private information for human body recognition in smart home. We propose a human body recognition scheme with the clustering based on the human body tracking using the sparse point cloud data of MIMO millimeter-wave radar. Firstly, the possible position of human body is predicted based on Kalman filter. Then, the point cloud data is clustered based on the human body shape in the prediction range of the human position. Finally, label tags are used to mark the human body targets detected by each frame of the radar. We apply human body recognition to validate the effectiveness of the proposed scheme. It can achieve single-person and double-person recognition using the sparse point cloud data of MIMO millimeter-wave radar. The results show that our proposed scheme reduces the error probability by 23.4% for the single-person recognition and by 31.1% for the double-person recognition. Extensive evaluations on the application of human activity recognition well demonstrate the practicability of the proposed scheme.