Sound detection and recognition based on biomimetic fiber-microcavity acoustic sensors

Inspired by the tympanum structures of insects, we present miniature biomimetic Fabry-Perot fiber-microcavity acoustic sensors. These sensors incorporate 3 types of biomimetic nano-structured diaphragms and achieve a maximum signal-to-noise ratio of 84.3 dB, allowing for the detection of pressure as low as 1.6 μPa/Hz^1/2. Moreover, the biomimetic fiber acoustic sensors exhibit varied spectral responses, offering a way to obtain more detailed sound information. By synchronizing the three sensor types and integrating them with neural networks, we develop high-accuracy acoustic recognition for human speech, achieving an identification error of less than 0.3%.