High-Precision, Low-Cost, and Low-Complexity Optical Touch Interface for Human–Computer Interaction on Arbitrary Surfaces

This study presents a high-precision, low-cost, and low-complexity optical touch interface designed for human–computer interaction (HCI) on arbitrary surfaces. The proposed system employs a dual-line CMOS linear image sensor configuration, enabling real-time multi-touch detection with minimal hardware overhead. By utilizing peak signal-to-noise ratio (PSNR) for touchpoint quality enhancement and a graph neural network (GNN) architecture for gesture classification, the system achieves accurate recognition of dynamic gestures and mid-air handwriting. Experimental validation across diverse environmental conditions demonstrates a positioning accuracy of 99.21%, sub-3.12 ms processing latency, and robust performance on edge AI devices. The integration of spatial and temporal attention mechanisms in the GNN further ensures high recognition fidelity while maintaining energy efficiency. This architecture supports plug-and-play deployment, making it an ideal solution for smart classrooms, digital whiteboards, and interactive displays in industrial or educational environments.