image: (a) Illustration of tactile challenges faced by Braille beginners. Overlapping receptive fields and strain diffusion produce blurred tactile information, often requiring repeated single‐point confirmation and slowing recognition. (b) The proposed iontronic array performs multi‐channel sensing during sliding input. With gradient‐modulus strain isolation to confine lateral strain transfer, the array preserves spatial fidelity and supports high‐resolution multi‐point recognition. The resulting temporal multi‐channel signals are processed through machine‐learning‐based decoding to identify Braille characters and generate auditory feedback, forming an assistive sensing workflow for Braille reading.
Credit: ©Science China Press
Under high-density tactile stimulation, the human somatosensory system is susceptible to strain diffusion within soft tissue and overlaps in receptive fields, leading to blurred initial spatial perception—an effect particularly pronounced among beginning Braille learners. At the engineering level, this phenomenon corresponds to mechanical crosstalk in high-density electronic skins induced by lateral strain transfer. Its cumulative effects directly degrade recognition stability during dynamic sliding and compromise long-term operational reliability. Addressing this critical challenge, this work introduces a low-crosstalk iontronic tactile array by incorporating a gradient high/low-modulus strain-isolation layer (G-HL-SIL) into the device architecture. This strategy effectively suppresses lateral strain propagation while maintaining high sensitivity, enabling stable, well-resolved multi-point spatial perception. When coupled with sliding-based signal acquisition and machine-learning-assisted decoding, the system achieves efficient Braille recognition with auditory feedback, providing a promising technological pathway for the reliable deployment of high-density electronic skins in assistive tactile sensing applications.
To uncover the mechanical origins of crosstalk in high-spatial-resolution tactile perception, a finite-element model of a 5×5-pixel array was constructed to systematically investigate how unit-area proportion (AP) and the modulus of the strain-isolation layer regulate strain propagation. The results reveal that reliable perception in high-density tactile arrays cannot be achieved through single-parameter optimization. Instead, coordinated design of AP and isolation-layer modulus is essential to achieving a robust balance between suppressing lateral strain diffusion and maintaining high sensitivity. These insights establish a quantitative structural design framework for next-generation high-density electronic skins.
The G-HL-SIL architecture enables decoupling of mechanical functionality along the thickness direction. The compliant lower region near the sensing interface preserves high normal compressibility to sustain sensitive pressure responses, while the upper high-modulus region constrains nonlocal strain propagation and forms a directional mechanical conduction pathway. Experimental results confirm that this configuration effectively localizes strain distribution under high-density conditions, significantly improving signal consistency and enhancing service stability under dynamic multi-point stimulation. These findings demonstrate both the effectiveness and scalability of the proposed structural strategy for large-area electronic skins.
Beyond suppressing interference at the single-pixel level, multi-bump recognition was further investigated to emulate realistic Braille reading conditions. Quantitative design boundaries for multi-point recognition in high-density tactile arrays were established, revealing unavoidable geometric constraints associated with excessively high AP or insufficient bump height. Based on national Braille standards, multi-character arrays were fabricated and achieved stable, reliable recognition performance within the validated low-crosstalk design window. The G-HL-SIL architecture accurately mapped Braille dot distributions, reduced false-trigger events by approximately 3.5-fold, and achieved 100% static and 99% dynamic recognition accuracy at the system level. Through multi-channel feature extraction combined with machine-learning-based classification, a “sliding–recognition–speech” platform was developed, demonstrating reading speed and reliability that significantly surpass those of beginning Braille learners.
Collectively, these results illustrate that structurally suppressing mechanical crosstalk not only directly improves recognition accuracy but also represents a powerful strategy for ensuring the long-term stable operation of high-density electronic skins. This work highlights a structure–material co-design paradigm that enables scalable, generalizable stability optimization in large-area tactile systems and provides an important technological foundation for future assistive perceptual technologies for visually impaired individuals.