Reducing Wind-Induced Noise in Millimeter-Wave Radar Point Cloud Data to Enhance Accuracy in Human Motion Monitoring

Millimeter-wave radar systems operate by transmitting and receiving electromagnetic signals within the millimeter-wave spectrum, utilizing the Doppler effect to determine the distance, velocity, and angle of target objects. Due to their high resolution and strong penetration capabilities over limited ranges, millimeter-wave radars are widely used for non-contact health monitoring. However, in indoor environments, wind can introduce noise into the point cloud signals, affecting the accuracy of human motion recognition. This study aims to mitigate wind-induced noise in millimeter-wave radar point cloud data. The research begins by simulating the impact of wind on human motion across diverse scenarios. By controlling wind volume and direction, point cloud data of humans in different states will be collected, and motion point cloud data in windless conditions will be gathered for comparative analysis. The point cloud data will then undergo preprocessing, including filtering, clustering, and voxel grid filtering methods. A transformer model will be employed to identify and attenuate wind noise features, with results compared against data obtained under windless conditions. Metrics such as recognition accuracy, changes in signal-to-noise ratio, and point cloud density variations will be recorded to evaluate the effectiveness of this method.