Integrating 3-D Deep Learning for Hand Point Cloud Segmentation to Enhance Biosensor-Based Human–Computer Interaction

Against the dual backdrop of normalized pandemic prevention and control measures and the intensification of population aging, the demand for noncontact hand function intelligent rehabilitation is becoming increasingly urgent. With the rapid development of virtual reality and human-computer interaction, accurate gesture posture perception is key to achieving intelligent rehabilitation systems. We propose a segmentation method based on 3-D deep learning and laser point clouds, where the hand point cloud segmentation network (HPCSN) can directly process point cloud data. The HPCSN innovatively integrates local aggregation, dimensional abstraction enhancement, and disorder handling of 3-D hand point cloud information, and has demonstrated excellent performance in comparative experiments with cutting-edge models. Our achievement is not only of great significance for improving the personalization and interest of rehabilitation training, but also provides a new technological path for hand function rehabilitation in an aging society, especially showing tremendous potential in remote medical care and intelligent monitoring applications. Our research provides a scientific basis for the future development of new types of biosensors and electronic devices, which is expected to be applied in various fields such as biomedical diagnosis and rehabilitation robots.