Flexible and Bridge-Free CuI Mutual Capacitive Arrays for Machine Learning-Assisted Three-Dimensional Spatial Interactive Interfaces

Flexible transparent human-machine interfaces (HMIs) are fundamentally bottlenecked by the brittleness of conventional indium tin oxide (ITO), the mechanical vulnerability of photolithographic bridged structures, and the complex hardware shielding required to suppress capacitive crosstalk. Herein, we propose a hardware-algorithm co-design strategy featuring a bridge-free double-sided (BFDS) mutual capacitive sensor array based on magnetron-sputtered p-type γ-CuI thin films. By employing a simplified floating-electrode readout scheme, the transparent flexible array not only captures high-fidelity two-dimensional (2D) tactile signals (signal-to-noise ratio of 268.5 ± 5.6) but also unlocks highly sensitive three-dimensional (3D) proximity perception (0-5 cm) driven by fringing electric fields. To overcome the global spatial crosstalk intrinsically induced by the unshielded hardware, a customized lightweight Convolutional Neural Network (CNN) is integrated. The CNN robustly decodes the complex nonlinear capacitance matrices, achieving an overall test accuracy of 86.2%. Furthermore, the CuI/PET structure exhibits good robustness over 10,000 extreme bending cycles (radius of 2.5 mm), where the grain-boundary scattering serves as a strain-release mechanism while preserving a continuous percolating conductive network. This work demonstrates that algorithmic compensation can relax stringent hardware shielding constraints, paving a novel avenue for ultra-simplified, touchless, and intelligent wearable electronics.