Machine-Learning-Assisted Fingerprint-Inspired Triboelectric Tactile Sensor for High-Performance Material and Texture Discrimination

The simultaneous and precise discrimination of material types and textural features constitutes a fundamental challenge in intelligent tactile perception owing to coupled inherent material charge transfer properties and texture-induced charge distributions. Inspired by human fingerprint biomechanics, this study presents a multimodal biomimetic fingerprint sensor that generates distinct polarized electric field responses at the contact interface under varying pressures and frequencies, enabling concurrent detection of material and texture characteristics. The sensor exhibits outstanding pressure sensitivity (3.41 kPa–1), facilitating precise material identification of planar specimens. Its superior texture perception capability enabled the accurate differentiation of smooth planar surfaces, hemispherical protrusions, conical protrusions, linear ridges with trapezoidal cross sections, and sinusoidal ridges with rectangular cross sections under controlled experimental conditions. A feature-decoupled random forest architecture was employed to address multimodal signal coupling across a data set comprising 821 experimental output voltage signals, enhancing interpretability while minimizing computational demands. The system achieved excellent signal linearity (R2 ≥ 0.95) through spatiotemporal feature extraction and adaptive weighting, demonstrating exceptional sensitivity and perfect identification accuracy for six visually similar disk-shaped tableware items. The proposed sensor provides a compact, high-performance solution for robotic tactile systems, intelligent prosthetics, and human–robot interactions while establishing a design paradigm for next-generation multimodal tactile perception via feature decoupling.