Abstract— Non-invasive brain machine interfaces (BMIs) on motor imagery movements have been widely studied and used for many years to take advantage of the intuitive link between imagined motor tasks and natural actions. En route to future technical applications of neuromorphic computing, a major current challenge lies in the identification and implementation of brain inspired algorithms to decode recorded signals. Neuro- morphic computing is believed to allow real-time implemen- tation of large scale spiking models for processing and compu- tation in non-invasive BMIs. Taking inspiration from the olfactory system of insects, we advance and implement a novel approach to decode and predict imaginary movements from electroencephalogram (EEG) signals. We use a spiking neural network implemented on SpiNNaker (4 chip, 64 cores) neuro- morphic hardware. Our work provides a proof of concept for a successful implementation of a functional spiking neural network for decoding two motor imagery (MI) movements on the SpiNNaker system. The approach can be extended to classify more complex MI movements on larger SpiNNaker systems. I. INTRODUCTION Neuromorphic platforms use a different architecture compared to sequential computing, one that mimics brain processing with neurons and synapses. These hardware systems achieve a speedup and support massively parallel data processing in a brain-inspired fashion. In contrast to expectations according to Moore's law, the integration density of digital processors ceased to grow during the last decade and is unlikely to increase again. Hence, neuromorphic architectures are perceived to be a potential solution for this issue, as they express fundamental differences when compared to conventional computing infrastructure [1]. To meet the challenge of efficient neuromorphic computing, a number of neuromorphic platforms have been and continue to be developed such as SpiNNaker, All authors are with the Technical University Munich, Germany, Department of Electrical and Computer Engineering, Research Group Neuroscientific System Theory, Center of Competence Neuroengineering. (+49 89 28926909, e-mail: zied.tayeb@tum.de) BrainScales hardware or IBM’s True North [2]. SpiNNaker is a novel massively-parallel computer architecture, inspired by the fundamental structure and function of the human brain. Each chip is a multi-core system, consisting of 18 ARM968- based cores and also several internetworking elements and supporting modules. Thus, it supports an emulation of up to 12,000 neurons in biologically plausible real-time per board (SpiNNaker-3) [3] or even about 200,000 neurons per board (SpiNNaker-5). This work presents a first attempt to use the potential advantages of neuromorphic computing and spiking neural network algorithms to decode brain signals and as per our knowledge no similar work has been published before. This new technology can enhance EEG based brain-machine interface classification performance which still suffer from limited motor imagery (MI) task detection and low accuracy due to non-stationary and non-linear characteristics of the brain signal. The aim of this work is therefore to decode EEG patterns generated during two MI movements (motor imagery of left and right hand) using a spiking neural network decoder. Here, as the first proof-of-principle, features of EEG signals were extracted and a classifier inspired by the olfactory system of insects was employed and implemented on SpiNNaker to recognize the two different MI movements. II. METHODS A. Data Description Our work is based on experimental data recorded by the Institute for Knowledge Discovery (Laboratory of Brain- Computer Interfaces), Graz University of Technology. EEG signals were collected from 9 subjects that were able to perform long and stable motor imagery over a minimum time of 2-s and data were sampled at 250Hz. More experimental details are provided in [4]. B. Preprocessing and Feature Extraction MI describes the mental rehearsal of a motor task without its execution, such as imagination of squeezing a training ball. These tasks induce a power increase or decrease of EEG amplitudes in certain frequency bands (mainly alpha and beta band) relative to a reference period and are referred Decoding of Motor Imagery Movements from EEG Signals using SpiNNaker Neuromorphic Hardware Zied Tayeb, Emeç Erçelik, and Jörg Conradt Neuroscientific System Theory, Technical University of Munich, 80333 München, Germany