Emulation of Auditory Nerve Adaptation with Bio-Inspired Acoustic Sensor to Extract Sound Features
Steve Durstewitz, Claudia Lenk, Tzvetan Ivanov, Martin Ziegler · 2024
Next generation acoustic sensor concepts use inspi-ration from biology. The advantages compared to traditional acoustic sensors are reduction of irrelevant or redundant data and adaptation of sensing properties to the environment. Since changing from quiet to noisy environment strongly impacts the recognition rate of sound analysis systems due to changes in the sound signals, adaptation to different environments is particularly important. Here we present a bio-inspired acoustic sensor that adapts to changes of sound signals and thereby mimics the adaptation at the auditory nerve in the inner ear. The bio-inspired acoustic sensor consists of a silicon beam with feedback loop and provides frequency decomposition and compressive amplification. To re-alize the adaptation, an analog circuit with discrete components was designed and realized on a PCB. The main idea of the presented circuit is to decrease the sensitivity of the sensor when the sound signal is constant. Thus, sound changes are highlighted and sound onset and offset can be extracted. Extracting such features, which are important for sound localisation and speech recognition, at sensor level can enable faster and more efficient sound analysis systems by reducing subsequent processing load and the energy consumption of the sensor system.