SpiNNak-Ear — Auditory pathway simulation on neuromorphic hardware
The first contribution of this work to the Speech In Noise (SpiN) workshop is to confuse matters by introducing a second ‘SpiN’ acronym. The SpiNNak-Ear system is an auditory pathway simulation tool implemented on the Spiking Neural Network architecture (SpiNNaker) neuromorphic platform.
A SpiNNaker machine is neuromorphic (brain inspired) in its design, where computational nodes (ARM microprocessors) are low power and spread across a vast network - much like networks of biological neurons. It was designed for applications of simulating large scale spiking neural networks; however, its massively parallel computational architecture is suitable for a range of applications that are not exclusively based around modelling spiking neurons.
SpiNNak-Ear takes advantage of the inherent parallel processing across the programmable computational nodes in the SpiNNaker system, achieving real-time binaural simulation of the early auditory pathway (from pinna to Auditory Nerve [AN]) to a biologically realistic scale (30,000 AN fibres).
Large scale simulation of neural cell populations in subsequent regions of brainstem nuclei are performed on the same SpiNNaker hardware. The scope for continuing this process into modelling the remainder of the binaural nuclei in the auditory pathway and associated cortical regions on this platform is achievable across a large scale machine of up to one million microprocessor cores.
The system presented here has the capabilities for modelling the auditory brain at a biologically realistic scale without incurring the inherent performance penalties of traditional computer architectures. It has advantages over alternative high-performance computational platforms by using a unique, brain inspired core-to-core ‘multi-cast’ communication protocol. This can be utilised when simulating the numerous inter-nuclei descending projections that feature in the auditory pathway.
The SpiNNak-Ear platform can aid future auditory research in explaining the complex in-vivo auditory neuron response, the development of future hearing prostheses, and in gaining a better understanding of the biological feedback pathways and their potential role in extracting salient stimuli from a noisy environment.