Digital cochlear modelling on FPGAs

Abstract

One should wonder about the complexity of the human body and understand the function of the organs in terms of mathematical control systems. Hearing is one of the senses of most mammals on earth. Such sense is available if the auditory system is functioning correctly. The functionality of the cochlea is to process non-linear travelling wave and give information to the human brain to perform tasks such as speech recognition and cross correlation of the incoming sound. In this dissertation, the functionality of the human cochlea was programmed on an FPGA board through in HDL. Such cochlear design consists of a cascade of IIR low-pass filters on each channel of the human auditory system; two in total, one for each ear. Each filter in the cascade is to be set to have a cut-off frequency set on an exponential distribution. The filter programming involved dual fixed point arithmetic which is has several advantages compared to normal fixed point arithmetic when used on FPGA boards and due limited hardware on the board used, processing in filter computations was carried using multiplexing procedures. Furthermore the IHC model and the AGC were designed to improve the functionality of the digital human cochlea.