From single- to multi-channel systems: Advancing handwriting forgery detection with triboelectric nanogenerator arrays

Abstract

Handwriting recognition is a critical tool in identity verification and document authentication, yet existing technologies face limitations such as susceptibility to forgery and dependency on professional expertise. In this study, we propose a multi-channel handwriting recognition system (MCHRS) based on triboelectric nanogenerators (TENG-Sensors) to address these challenges. The system integrates a TENG-based handwriting tablet (TENG-HT) with deep learning and an OC-SVM classifier for accurate and efficient handwriting recognition. The TENG-Sensors generate distinct voltage signals during handwriting, capturing dynamic pressure information unique to each character. We systematically evaluated the detection accuracy of TENG-HTs with 1, 2, and 4 channels, demonstrating that the 4-channel configuration achieved the highest recognition accuracy. Using the MobileNet V2 model for feature extraction, the system accurately distinguished between handwriting by genuine writers and forgers. Additionally, the MCHRS was enhanced with wireless data transmission capabilities through integration with ADC, MCU, and WiFi modules, enabling real-time processing without external power supply. The results highlight the superior performance of the 4-channel MCHRS, achieving over 99 % recognition accuracy in distinguishing handwritten Chinese and numeric characters. This self-powered, wireless system demonstrates significant potential for practical applications in handwriting recognition, offering a robust, cost-effective, and forgery-resistant solution.