A Bio-Inspired Cochlear Heterodyning Architecture for an RF Fovea

We discuss the use of cochlear models for spectrum analysis at radio frequencies. We describe performance characteristics of such models, including noise, dynamic range, and frequency resolution. We show that the addition of phase information improves frequency estimation as compared to the use of amplitude information alone. In particular, the use of both amplitude and phase information in a novel nonlinear bio-inspired center-surround coincidence-detection stage simultaneously improves frequency estimation and implements a lowpass-to-bandpass transformation on cochlear outputs. In order to further improve frequency estimation we propose a novel wireless receiver architecture that is a broadband generalization of narrowband heterodyning systems commonly used in radio. We term this architecture cochlear heterodyning. It exploits the efficiency of cochlear spectrum analysis to perform parallel, multi-scale analysis of wideband signals and can be constructed with cochlea-like traveling-wave structures. When combined with our prior work on an RF cochlea, such architectures may be useful in cognitive radios for creating RF foveas that select narrowband components present within wideband, but spectrally sparse signals. The operation of RF foveas is analogous to how the eye foveates on narrow but interesting portions of an image. Analogies between spectrum analysis and the process of successive-subranging analog-to-digital conversion illustrate how successively finer frequency resolution is achieved in an RF fovea. Finally, we show that RF foveas can be used in feedback loops to perform interference cancellation.