Toward neuromorphic intelligent brain-machine interfaces: An event-based neural recording and processing system

Abstract

We present an analog neural recording front-end design that can be easily interfaced with Address-Event Representation (AER) neuromorphic systems via an asynchronous digital communication channel. The proposed circuits include a low-noise amplifier for biological signals, a delta-modulator analog-to-digital converter, and a low-power bandpass filter. The bio-amplifier has a gain of 54 dB, with an Root Mean Squared (RMS) input-referred noise level of 2.1 μV, and consumes 90 μW. The bandpass filter and delta-modulator circuits include asynchronous handshaking interface logic compatible with the AER communication protocol. We describe the circuits, present experimental measurements to demonstrate their response properties and show how they can be used in conjunction with neuromorphic computing architectures to implement decoding and learning functions useful for Brain-Machince Interfaces (BMIs).