CAR-Lite: A Multi-Rate Cochlear Model on FPGA for Spike-Based Sound Encoding

Filters in cochlear models use different coefficients to break sound into a 2-D time-frequency representation. On digital hardware with a single sampling rate, the number of bits required to represent these coefficients requires substantial computational resources such as memory storage. In this paper, we present a cochlear model operating at multiple sampling rates. As a result, fewer bits are required to represent filter coefficients on hardware as opposed to all the filters operating at a single sampling rate; with a 108-filter cochlear implementation, up to nine times fewer coefficients are needed. We present an implementation of this model in Matlab and on an Altera Cyclone V field-programmable gate array. We also demonstrate the capability of our model to encode sound at various intensity levels and with real-world signals.