Evolvable Analogue Electronic Artificial Cochlea using Field Programmable Analogue Array

An evolvable analogue electronic artificial cochlea has become a practical proposition with the recent development of dynamically configurable and reconfigurable Field Programmable Analogue Arrays (FPAA). The authors describe a proof-of-concept system based on the Anadigm® FPAA that implements a four-channel evolvable cochlea on one device. A genetic algorithm (GA) can be used to arrive at an optimal set of filter parameters in a given environment. In this way, the system is able to adapt the filter bank for different acoustic environments. The system proposed learns to adapt to the acoustic environment and tunes itself to the wearers needs in that environment. The authors are also investigating how the system could be further refined to evolve the configuration of the cochlea through the primary configuration of the FPAA. This might include, for example, varying the number of channels, comparing multi-channel to single channel, pulse (spiking) versus analogue signal waveform or by using more powerful evolutionary algorithms than the GA used such as the harmony search algorithm.