Flexible Pressure Sensor Array with Mortise‐and‐Tenon Interlocking Structure for Deep Learning‐Assisted Keyboard Input and Handwriting Recognition

Abstract The integration of flexible sensors and deep learning has significantly advanced human‐machine interaction. However, developing high‐resolution and cost‐effective pressure sensor arrays (PSAs) remains a challenge. This study presents a PSA with a mortise‐and‐tenon interlocking structure, composed of four T‐shaped capacitive pressure sensors (TCPSs). The TCPS is fabricated using low‐cost commercial materials such as conductive polyurethane foam, polyester fabric, and polyimide film, and demonstrates region‐specific sensitivity, rapid response time, and excellent stability. When assembled into a mortise‐and‐tenon interlocking sensor array (MTIPSA), the spatial resolution improves to over 200% of that of conventional designs. The MTIPSA excels in flexible interaction applications: as a flexible keyboard, it achieves 99.33% accuracy in classifying 30 types of inputs (letters and punctuation) using a convolutional neural network (CNN), and as a writing pad, it recognizes digits (0–9) with 99.25% accuracy. By integrating a CNN with a long short‐term memory (LSTM) network enhanced by a self‐attention mechanism (SAM), the system achieves 99.5% accuracy in identifying individual users from handwriting samples. This work has broad application prospects in smart electronic skin and human–machine interaction.