The biosensorics group is engaged in the development of bio-inspired smart acoustic sensors based on silicon microresonators. The goal is an adaptive, nonlinear sensor system,
- whose properties (e.g. signal-to-noise ratio, bandwidth, dynamic range) can be specifically adapted to the respective requirements, or which can adapt itself to them
- which mainly or only records relevant data, so that subsequent filtering or processing of the signals to separate relevant data (such as speech) from irrelevant data (such as noise) is no longer required
- which allows a pre-processing of the data (e.g. frequency decomposition, selective and non-linear amplification, sound or speech recognition) so that the data flow to the following signal processing units is reduced, the signal processing (e.g. speech recognition) is improved or the signal processing is directly integrated into the sensor
- whose energy efficiency is significantly lower than that of conventional speech recognition/audio signal processing devices comprising microphones, filters, amplifiers and signal processing elements.
To realize bio-inspired, smart acoustic sensors based on silicon microresonators we use effects of nonlinear dynamics, feedback loops, the combination of the sensor system with neuromorphic electronics (e.g. memristors, but also neuromorphic neural networks) and the principles of neuromorphic information processing.The development is based on the one hand on the underlying mechanisms of biological sound detection in the inner ear (cochlea) and on the other hand on the principles of neuromorphic information processing/artificial intelligence.
We see possible application areas of the sensor system e.g. in:
- hearing aids or cochlear implants for the treatment of hearing disorders as biomimetic microphones
- speech-controlled devices
- speech processing devices