Intelligent Song Recognition via a Hollow‐Microstructure‐Based, Ultrasensitive Artificial Eardrum
Shaopeng Li, Jiangtao Tian, Ke Li, Kemeng Xu, Jiaqi Zhang, Tingting Chen, Yang Li, Hongbo Wang, Qiye Wu, Jinchun Xie, Yongjun Men, Weiping Liu, Xiaodan Zhang, Wenhan Cao, Zhongjie Huang
Abstract
Abstract Artificial ears with intelligence, which can sensitively detect sound—a variant of pressure—and generate consciousness and logical decision‐making abilities, hold great promise to transform life. However, despite the emerging flexible sensors for sound detection, most success is limited to very simple phonemes, such as a couple of letters or words, probably due to the lack of device sensitivity and capability. Herein, the construction of ultrasensitive artificial eardrums enabling intelligent song recognition is reported. This strategy employs novel geometric engineering of sensing units in the soft microstructure array (to significantly reduce effective modulus) along with complex song recognition exploration leveraging machine learning algorithms. Unprecedented pressure sensitivity (6.9 × 10 3 kPa −1 ) is demonstrated in a sensor with a hollow pyramid architecture with porous slants. The integrated device exhibits unparalleled (exceeding by 1–2 orders of magnitude compared with reported benchmark samples) sound detection sensitivity, and can accurately identify 100% (for training set) and 97.7% (for test set) of a database of the segments from 77 songs varying in language, style, and singer. Overall, the results highlight the outstanding performance of the hollow‐microstructure‐based sensor, indicating its potential applications in human–machine interaction and wearable acoustical technologies.