mmPencil: Toward Writing-Style-Independent In-Air Handwriting Recognition via mmWave Radar and Large Vision-Language Model

With the rapid advancement of wireless technologies, in-air handwriting recognition based on radio frequency (RF) signals emerges as a promising solution for human-computer interaction. However, existing methods often constrain writing gestures to predefined planes, impose strict requirements on writing order and direction, and limit recognition tasks to a small set of digits or letters. To overcome these limitations, we propose m2VLMs, a modality mapping architecture, which serves as a bridge between the millimeter-wave (mmWave) radar sensing technologies and large vision-language models (VLMs). Based on the proposed architecture, a three-dimensional (3D) in-air handwriting word recognition system named mmPencil is developed. Specifically, we design a multi-stage spatial trajectory reconstruction algorithm that extracts frequency-domain features to identify handwriting regions and achieve high-precision reconstruction of 3D word trajectories. Furthermore, we introduce a novel spatial-to-visual mapping algorithm, which bridges the gap between spatial trajectory information captured by mmWave radar and vision-language representations, providing a foundation for cross-modality understanding in the large vision-language model. As a result, mmPencil tackles the limitations of current solutions that are heavily reliant on handwriting styles and environmental factors, expanding RF-based in-air handwriting recognition to more complex word-level scenarios. We collect and release a 3D mmWave handwriting dataset comprising 200 distinct words (ranging from 2 to 9 letters), contributions from 12 users, and 22 different writing scenarios, totaling 7,664 samples with an overall size of 31.66 GB. Extensive experiments demonstrate that mmPencil achieves accurate and robust word recognition in real-world scenarios, remaining unaffected by variations in word categories, length, as well as writing position, range, angle, speed, size, direction, and user movement. Specifically, for 4 seen users, mmPencil achieves a recognition accuracy of approximately 97.60% across 200 word classes. Moreover, benefiting from the generalization capability of VLMs, the zero-shot recognition accuracy for 4 unseen users can also reach 92.50% across 50 word classes using training data from only 8 users, which outperforming state-of-the-art baselines.