IEEE Trans Biomed Circuits Syst. 2019 Dec;13(6):1264-1276. doi: 10.1109/TBCAS.2019.2948326. Epub 2019 Oct 21.
Tactile sensing requires form-fitting and dense sensor arrays over large-areas. Hybrid systems, combining Large-Area Electronics (LAE) and silicon-CMOS ICs to respectively provide diverse sensing and high-performance computation/control, enable a platform for such sensing. A key challenge is that hybrid systems require a large number of interfaces between the LAE and CMOS domains, particularly as the number of sensors scales. This paper presents an architecture that exploits the attribute of signal sparsity, commonly exhibited in large-scale tactile-sensing applications, to reduce the interfacing complexity to a level set by the sparsity rather than the number of sensors. This enhances scalability compared to sequential-scanning and active-matrix approaches. The architecture implements compressed sensing via thin-film-transistor (TFT) switches, and is demonstrated in a force-sensing system with 20 force sensors, a TFT die (with 161 ZnO TFTs) per sensor, and a custom CMOS IC for system readout and control. Acquisition error of 0.7 k[Formula: see text] is achieved over a 100 k Ω-20 k Ω sensing range, at energy and rate of 2.46 μ J/frame and 31 fps.
触觉感应需要贴合大面积的密集传感器阵列。混合系统结合了大面积电子学（LAE）和硅 CMOS IC，分别提供多样化的感应和高性能计算/控制，为这种感应提供了一个平台。一个关键的挑战是，混合系统需要在 LAE 和 CMOS 域之间有大量的接口，特别是当传感器数量增加时。本文提出了一种架构，利用在大规模触觉感应应用中常见的信号稀疏性的特性，将接口复杂度降低到由稀疏性而不是传感器数量决定的水平。与顺序扫描和有源矩阵方法相比，这种架构提高了可扩展性。该架构通过薄膜晶体管（TFT）开关实现压缩感知，并在一个具有 20 个力传感器的力感测系统中得到了验证，每个传感器有一个 TFT 管芯（具有 161 个 ZnO TFT）和一个用于系统读取和控制的定制 CMOS IC。在 100 kΩ-20 kΩ的感应范围内，实现了 0.7 k[Formula: see text]的采集误差，能量和速率分别为 2.46 μ J/帧和 31 fps。
