72 IEEE TRANSACTIONS ON BIOMEDICAL CIRCUITS AND SYSTEMS, VOL. 9, NO. 1, FEBRUARY 2015 per Channel Analog Biomimetic Cochlear Implant Processor Filterbank Architecture With Across Channels AGC Guang Yang, Richard F. Lyon, Fellow, IEEE, and Emmanuel. M. Drakakis, Member, IEEE Abstract—A new analog cochlear implant processor filterbank architecture of increased biofidelity, enhanced across-channel contrast and very low power consumption has been designed and prototyped. Each channel implements a biomimetic, asymmetric bandpass-like One-Zero-Gammatone-Filter (OZGF) transfer function, using class-AB log-domain techniques. Each channel's quality factor and suppression are controlled by means of a new low power Automatic Gain Control (AGC) scheme which is coupled across the neighboring channels and emulates lateral inhibition (LI) phenomena in the auditory system. Detailed mea- surements from a five-channel silicon IC prototype fabricated in a AMS technology confirm the operation of the coupled AGC scheme and its ability to enhance contrast among channel outputs. The prototype is characterized by an input dynamic range of 92 dB while consuming only of power in total ( per channel) under a 1.8 V power supply. The architec- ture is well-suited for fully-implantable cochlear implants. Index Terms—Analog integrated circuits, automatic gain con- trol, cochlear implant, lateral inhibition, low power, spectral en- hancement. I. INTRODUCTION O VER the past 30 years, cochlear implants (CIs) have de- veloped from a device that was thought impossible for speech recognition and useful only for sound perception to an established clinical device for restoring partial hearing to deaf people. In the near future CIs will become fully implantable: all the external components of a CI system (e.g., the microphone and its speech processor) will be implanted except for a remote controller required to be external for the programming of the implanted part. As a result, CI users will be indistinguishable in appearance from normal hearing people, which can boost their self-confidence and improve third-party attitudes to them; such improvements have already been witnessed with state-of-the-art hearing aids which are placed inside the ear canal and are thus invisible or nearly invisible. Other benefits include: 1) fewer Manuscript received October 21, 2013; revised February 14, 2014; accepted April 24, 2014. Date of publication July 23, 2014; date of current version Jan- uary 23, 2015. This paper was recommended by Associate Editor S.-C. Liu. G. Yang was with the Department of Bioengineering, Imperial College London, London SW7 2AZ, U.K. He is now with the Department of Electrical and Electronic Engineering, University of Bristol, Bristol BS8 1TR, U.K. (e-mail: yangsihai84@hotmail.com). R. F. Lyon is with Google Inc., Mountain View, CA 94043 USA (e-mail: dicklyon@acm.org). E. M. Drakakis is with the Department of Bioengineering, Imperial College London, London SW7 2AZ, U.K. (e-mail: e.drakakis@imperial.ac.uk). Digital Object Identifier 10.1109/TBCAS.2014.2325907 limitations to CI users' daily activities, 2) restoring access to the directional amplification function of the pinna by means of the microphone placed inside the ear canal, and 3) eliminating the constraints in the data bandwidth of the RF link transmitting the processed microphone input to the implanted electrodes. The filterbank architecture presented in this work, termed “OZGF-with-LI”, is intended for use in such CI systems. Each channel implements the One-Zero-Gammatone-Filter (OZGF) [1] while the novel automatic gain control (AGC) scheme, coupled across channels, emulates lateral inhibition (LI) phe- nomena in the auditory system. The OZGF was adopted as it can provide a good model of auditory filtering with only three parameters [1], [2] while the LI is regarded as an important biological spectral-enhancement mechanism which may partly account for the high robustness of the normal auditory system to noise [3]–[6]. The architecture performs multi-channel syllabic compression while preserving well across-channel contrast and hence the original spectral features of the system input. This is an advantage over the compression schemes used in current CIs, which tend to degrade spectral contrast through an asymmetric amplification over channel-specific frequency regions (i.e., the valleys in the spectrum and the corresponding weak-amplitude channels are strongly amplified in relative terms so that they are concurrently audible with weakly amplified spectral peaks and the corresponding “strong” channels; this side effect reduces the differences between channel outputs) [7]–[9]. This paper details a novel ultra-low-power analog VLSI implementation of this filterbank, explicitly designed to meet stringent power requirements compatible with fully implantable CI processors. Our filterbank does not intend to replicate the “exact” biological operations of the auditory system, which can be a complex and power hungry task [10]; instead, it aims to achieve a good trade-off between bio-fidelity and power con- sumption. Our implementation opts for an analogue solution rather than its digital counterpart because the former is known to provide considerable saving in both power consumption and silicon area compared to the latter when the precision required at the output is low [11]: typically, a CI processor's channel bandwidth is a few kHz at most and a patient's dynamic range (DR) is only 3–20 dB. In the proposed implementation, the input DR together with the power consumption of each channel were optimized with micropower companding techniques [12]–[20] including log- domain filters biased dynamically via the AGC (syllabic com- panding). For each channel the biasing of the log-domain filter stages varies adaptively according to a coupled and weighted 1932-4545 © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.