In‐Sensor Touch Analysis for Intent Recognition

• Published in: Advanced functional materials 2024-09, Vol.34 (52), p.n/a
• Main Authors: Xu, Yijing, Yu, Shifan, Liu, Lei, Lin, Wansheng, Cao, Zhicheng, Hu, Yu, Duan, Jiming, Huang, Zijian, Wei, Chao, Guo, Ziquan, Wu, Tingzhu, Chen, Zhong, Liao, Qingliang, Zheng, Yuanjin, Liao, Xinqin
• Format: Article
• Language: English
• Subjects: bionic structure; Computation; human–machine interactions; intent recognitions; in‐sensor computing; Sensor arrays; Sensors; Touch; touch sensors
• ISSN: 1616-301X; 1616-3028
• DOI: 10.1002/adfm.202411331

Tactile intent recognition systems, which are highly desired to satisfy human's needs and humanized services, shall be accurately understanding and identifying human's intent. They generally utilize time‐driven sensor arrays to achieve high spatiotemporal resolution, however, which encounter inevitable challenges of low scalability, huge data volumes, and complex processing. Here, an event‐driven intent recognition touch sensor (IR touch sensor) with in‐sensor computing capability is presented. The merit of event‐driven and in‐sensor computing enables the IR touch sensor to achieve ultrahigh resolution and obtain complete intent information with intrinsic concise data. It achieves critical signal extraction of action trajectories with a rapid response time of 0.4 ms and excellent durability of >10 000 cycles, bringing an important breakthrough of tactile intent recognition. Versatile applications prove the integrated functions of the IR touch sensor for great interactive potential in all‐weather environments regardless of shading, dynamics, darkness, and noise. Unconscious and even hidden action features can be perfectly extracted with the ultrahigh recognition accuracy of 98.4% for intent recognition. The further auxiliary diagnostic test demonstrates the practicability of the IR touch sensor in telemedicine palpation and therapy. This groundbreaking integration of sensing, data reduction, and ultrahigh‐accuracy recognition will propel the leapfrog development for conscious machine intelligence. An event‐driven intent recognition touch sensor composed of the biomimetic helical web with in‐sensor computing capability is developed to achieve high resolution and obtain complete intent information while remaining concise data. It achieves the successful extraction of action trajectories with a rapid response time of 0.4 ms and excellent durability of >10 000 cycles, bringing an important breakthrough of tactile intent recognition.