Bio-inspired acoustic sensor with gain adaptation enhancing dynamic range and onset detection

The exceptional capability of humans to perceive speech in varying acoustic environments could not yet be achieved by technological speech processing systems. Possible reasons for this might be that (i) signal pre-processing occurring in the inner ear before and during transduction is implemented in technological systems only after the transduction stage and (ii) adaptation of the transduction stage with incorporated pre-processing is not included. Here, we present a novel, nonlinear acoustic sensor system including adaptation. Thereby, (short-term) adaptation is implemented using automatic gain control by discrete analog circuits. Thus, the sensor includes pre-processing capabilities like frequency decomposition, nonlinear amplification of signals and adaptation of sensing properties to varying acoustic inputs. The integration of the adaptation mechanism into the sensor system not only enhances important sound features like sound onset but also increases its dynamic range by increasing sensitivity for low volume sounds and preventing saturation for large volume sounds. A flexible design with tuneable components allows adaptation on different time scales (short- and long-term adaptation).