Multilevel nanostructured pressure sensor for object recognition with Deep-Learning assistance: A strategy for high sensitivity and wide detection range

Abstract

Highly sensitive, flexible pressure sensors with a broad range of pressure responses are highly desired in the haptic recognition application. However, achieving a substantial broad sensing range while offering nanoscale-sensitive response presents a formidable challenge. To resolve this, we created a multilevel nanostructure inspired by the natural structure of cicada wings, with carefully designed back-to-back structural layout and optimized material adhesion using a non-uniform spray. The composite material was assembled in a pressure sensor with a broad pressure detection range from 0 to 500 kPa with a remarkable full-range sensitivity of 16792 kPa−1, making it ideal for full-scale human activity monitoring. Combined with the deep–learning algorithm, the constructed flexible multiplex 2D-array pressure sensors effectively recognized objects with similar shapes, achieving a recognition accuracy of up to 98 %. This work presented a novel biomimetic multilevel surface structure for overcoming the limit of current pressure sensors, broadening its applications in wearable electronic sensing devices.