Speaker
Description
Bio-inspired acoustic sensors, which integrate functionalities of the human cochlea like frequency decomposition and non-linear amplification directly in the sensor itself, offer a great potential to improve speech-in-noise processing and energy-efficiency for embedded systems and hearing aids. Particularly, the dynamic MEMS cochlea demonstrated advantages regarding robustness of sound recognition in noisy conditions and enables adaptation of sensing and processing capabilities to the current hearing situation. It is based on micro-electromechanical silicon cantilevers in combination with electronic feedback. To cover the auditory frequency range, an array of MEMS cochlea sensors is required due to their frequency filtering properties. This requires miniaturization of the sensor system and electronics to take advantage of its positive properties in next-generation hearing aids.Here, we present the design and fabrication of an array of 8 sensors. Each of these sensors responds to a different frequency, dependent on their geometry. However, due to the mechanical and acoustical coupling between them, artefacts can occur. To optimise the design regarding acoustic sensitivity, detect coupling artefacts and derive sensor parameters for simplified descriptions of the system, FEM simulations are performed and compared to measurements of different arrays, using different actuation schemes for the array, as well as results from the simplified single mode ODE model. Based on these results, the combination of multiple 8-sensor arrays can be designed for covering the auditory range and improve speech processing and sound quality.