Point Cloud Classification Fall Detection Method Based on Millimeter Wave Radar

Falling at home is the most common cause of injury among the elderly. Wearable devices are inconvenient and costly, while camera systems pose risks to privacy. A fall detection method employing millimeter-wave radar coupled with a classification network is proposed to address these issues. This approach leverages the benefits of millimeter-wave radar and deep learning. The raw data processed by the radar yields the required information, which is then rendered into a point cloud. Subsequent voxelization enhances feature learning, enabling the model to function even with sparse data. The information is then fed into our custom-designed network for time series extraction to classify behaviors. Finally, the time series extraction network utilized in this study is compared with several popular networks, demonstrating that our system offers an effective technological avenue for target detection and posture analysis, and boasts high algorithmic precision in millimeter-wave radar fall detection.