Domain-Independent Handwriting Trajectory Recovery Based on Multiple Wireless Links

Leveraging wireless signals for device-free sensing and handwriting trajectory recovery has enormous potential in human-machine interaction applications. However, wireless channel characteristics vary with the position and orientation of the user, as well as the locations of transmitter and receiver. This leads to serious performance degradation of common sensing and recognition algorithms when the user changes its position or orientation. The phenomenon is known as domain-dependency problem, which is a great challenge for the practical application of various wireless sensing schemes. This paper introduces a domain-independent handwriting trajectory recovery scheme based on multiple wireless links. By using a preamble gesture to extract the domain information, we can convert the Doppler frequency shifts of different transmitter-receiver pairs to velocities in the body coordinate system, without any prior location information. The high-precision 3D handwriting trajectory can then be recovered, and downstream tasks such as gesture recognition can be benefited. Based on the matrix perturbation theory, the error propagation in the trajectory recovery process is studied, and the impacts of geometric layout and preamble gesture are thoroughly analyzed. To validate the proposed approach, we conduct experiments with 5G-NR and Wi-Fi signals, and demonstrate the recovered trajectories under various configurations. In addition, based on an open-source dataset, we test the performance of handwriting recognition and achieve better accuracy than the state-of-the-art approach.