The MNIST dataset has become a standard benchmark for learning, classification and computer vision systems. Contributing to its widespread adoption are the understandable intuitive nature of task, relatively small size storage requirements accessibility ease-of-use database itself. was derived from larger known as NIST Special Database 19 which contains digits, uppercase lowercase handwritten letters. This paper introduces variant full dataset, we have called Extended (EMNIST), follows same...

Creating datasets for Neuromorphic Vision is a challenging task. A lack of available recordings from sensors means that data must typically be recorded specifically dataset creation rather than collecting and labeling existing data. The task further complicated by desire to simultaneously provide traditional frame-based allow direct comparison with Computer algorithms. Here we propose method converting static image into using an actuated pan-tilt camera platform. Moving the sensor scene or...

The MNIST dataset has become a standard benchmark for learning, classification and computer vision systems. Contributing to its widespread adoption are the understandable intuitive nature of task, relatively small size storage requirements accessibility ease-of-use database itself. was derived from larger known as NIST Special Database 19 which contains digits, uppercase lowercase handwritten letters. This paper introduces variant full dataset, we have called Extended (EMNIST), follows same...

We present an FPGA implementation of a re-configurable, polychronous spiking neural network with large capacity for spatial-temporal patterns. The proposed generates delay paths de novo, so that only connections actually appear in the training patterns will be created. This allows to use all axons (variables) store information. Spike Timing Dependent Delay Plasticity is used fine-tune and add dynamics network. time multiplexing approach allowing us achieve 4096 (4k) neurons up 1.15 million...

The advent of large scale neural computational platforms has highlighted the lack algorithms for synthesis structures to perform predefined cognitive tasks. Neural Engineering Framework offers one such synthesis, but it is most effective a spike rate representation information, and requires number neurons implement simple functions. We describe network method that generates synaptic connectivity which process time-encoded signals, makes very sparse use neurons. allows user specify –...

The growing demands placed upon the field of computer vision have renewed focus on alternative visual scene representations and processing paradigms. Silicon retinea provide an means imaging environment, produce frame-free spatio-temporal data. This paper presents investigation into event-based digit classification using N-MNIST, a neuromorphic dataset created with silicon retina, Synaptic Kernel Inverse Method (SKIM), learning method based principles dendritic computation. As this work...

Brain waves are rhythmic voltage oscillations emerging from the synchronization of individual neurons into a neuronal network. These range slow to fast fluctuations, and classified by power frequency band, with different bands being associated specific behaviours. It has been postulated that at least ten distinct mechanisms required cover neural oscillations, however gear transition between oscillatory frequencies unknown. In this study, we have used electrophysiological recordings explore...

We present a hardware architecture that uses the neural engineering framework (NEF) to implement large-scale networks on field programmable gate arrays (FPGAs) for performing massively parallel real-time pattern recognition. NEF is capable of synthesising cognitive systems from subnetworks and we have previously presented an FPGA implementation successfully performs nonlinear mathematical computations. That work was developed based compact digital core, which consists 64 neurons are...

In this work, we present an optical space imaging dataset using a range of event-based neuromorphic vision sensors. The unique method operation sensors makes them ideal for situational awareness (SSA) applications due to the sparseness inherent in data. These offer significantly lower bandwidth and power requirements making particularly well suited use remote locations space-based platforms. We first publicly-accessible including recordings from multiple providers, greatly lowering barrier...

Unsupervised feature extraction algorithms form one of the most important building blocks in machine learning systems. These are often adapted to event-based domain perform online neuromorphic hardware. However, not designed for purpose, such typically require significant simplification during implementation meet hardware constraints, creating trade offs with performance. Furthermore, conventional generate useful intermediary signals which valuable only context limitations. In this work a...

It has been more than two decades since the first neuromorphic Dynamic Vision Sensor (DVS) sensor was invented, and many subsequent prototypes have built with a wide spectrum of applications in mind. Competing against state-of-the-art neural networks terms accuracy is difficult, although there are clear opportunities to outperform conventional approaches power consumption processing speed. As sensors generate sparse data at focal plane itself, they inherently energy-efficient, data-driven,...

In this work, we investigate event-based feature extraction through a rigorous framework of testing. We test hardware efficient variant Spike Timing Dependent Plasticity (STDP) on range spatio-temporal kernels with different surface decaying methods, decay functions, receptive field sizes, numbers, and back end classifiers. This detailed investigation can provide helpful insights rules thumb for performance vs. complexity trade-offs in more generalized networks, especially the context...

As the interest in event-based vision sensors for mobile and aerial applications grows, there is an increasing need high-speed highly robust algorithms performing visual tasks using data. event rate network structure have a direct impact on power consumed by such systems, it important to explore efficiency of encoding used these sensors. The work presented this paper represents first study solely focused effects both spatial temporal downsampling data makes use variety sets chosen fully...

Abstract As an emerging approach to space situational awareness and imaging, the practical use of event-based camera (EBC) in imaging for precise source analysis is still its infancy. The nature data collection needs be further explored develop more effective systems advance capabilities tracking with improved target measurement models. Moreover, event measurements meaningful, a framework must investigated EBC calibration project events from pixel array coordinates image plane resident...

The Event-Based Sensor (EBS) is a new class of imaging sensor where each pixel independently reports "events" in response to changes log intensity, rather than outputting image frames containing the absolute intensity at pixel. Positive and negative events are emitted from when change exceeds certain controllable thresholds internal device. For objects moving through field view, can be related motion. records asynchronously for with very high temporal resolution, allowing detection quickly...

We present an end-to-end trainable modular event-driven neural architecture that uses local synaptic and threshold adaptation rules to perform transformations between arbitrary spatio-temporal spike patterns. The represents a highly abstracted model of existing Spiking Neural Network (SNN) architectures. proposed Optimized Deep Event-driven network Architecture (ODESA) can simultaneously learn hierarchical features at multiple time scales. ODESA performs online learning without the use error...

We report on the Falcon neuro event-based sensor (EBS) instrument that is designed to acquire data from lightning and sprite phenomena currently operating International Space Station. The consists of two independent, identical EBS cameras pointing in fixed directions, toward nominal forward direction flight Nadir direction. payload employs stock DAVIS 240C focal plane arrays along with custom-built control readout electronics remotely interface cameras. To predict sensor’s ability...

Earth orbit is a limited natural resource that hosts vast range of vital space-based systems support the international community's national, commercial and defence interests. This rapidly becoming depleted with over-crowding in high demand orbital slots growing presence space debris. We propose Fast Iterative Extraction Salient targets for Tracking Asynchronously (FIESTA) algorithm as robust, real-time reactive approach to optical Space Situational Awareness (SSA) using Event-Based Cameras...

Peripheral neuropathic desensitization associated with aging, diabetes, alcoholism and HIV/AIDS, affects tens of millions people worldwide, there is little or no treatment available to improve sensory function. Recent studies that apply imperceptible continuous vibration electrical stimulation have shown promise in improving sensitivity both diseased healthy participants. This class interventions only has an effect during application, necessitating the design a wearable device for everyday...

An event-based image sensor works dramatically differently from the conventional frame-based sensors in a way that it only responds to local brightness changes whereas its counterparts’ output is linear representation of illumination over fixed exposure time. The an therefore asynchronous stream spatial-temporal events data tagged with location, timestamp and polarity triggered events. Compared traditional sensors, have advantages high temporal resolution, low latency, dynamic range power...