A5079540787- Advanced Memory and Neural Computing
- CCD and CMOS Imaging Sensors
- Neuroscience and Neural Engineering
- Neural dynamics and brain function
- Analog and Mixed-Signal Circuit Design
- Neural Networks and Applications
- Advancements in Semiconductor Devices and Circuit Design
- Ferroelectric and Negative Capacitance Devices
- Low-power high-performance VLSI design
- Photoreceptor and optogenetics research
- Neural Networks and Reservoir Computing
- VLSI and Analog Circuit Testing
- Advancements in PLL and VCO Technologies
- Semiconductor materials and devices
- VLSI and FPGA Design Techniques
- Neural Networks Stability and Synchronization
- EEG and Brain-Computer Interfaces
- Blind Source Separation Techniques
- Radio Frequency Integrated Circuit Design
- Error Correcting Code Techniques
- Advanced Vision and Imaging
- Modular Robots and Swarm Intelligence
- Quantum-Dot Cellular Automata
- Embedded Systems Design Techniques
- Personal Information Management and User Behavior
Instituto de Microelectrónica de Sevilla
2016-2025
Universidad de Sevilla
2016-2025
Consejo Superior de Investigaciones Científicas
2009-2025
Politecnico di Milano
2022
Centro Nacional de Microelectrónica
1998-2021
AGH University of Krakow
2017
National Research Council
2008
Hardware implementations of spiking neurons can be extremely useful for a large variety applications, ranging from high-speed modeling large-scale neural systems to real-time behaving systems, bidirectional brain-machine interfaces. The specific circuit solutions used implement silicon depend on the application requirements. In this paper we describe most common building blocks and techniques these circuits, present an overview wide range neuromorphic neurons, which different computational...
In this paper we present a very exciting overlap between emergent nano technology and neuroscience. We are linking one type of memristor devices to the biological synaptic update rule known as Spike-Time-Dependent-Plasticity found in real synapses. Understanding link allows neuromorphic engineers develop circuit architectures that use memristors artificially emulate parts visual cortex. focus on referred voltage driven our discussions behavioral macro model for such devices. The...
In this paper we review several ways of realizing asynchronous Spike-Timing Dependent Plasticity (STDP) using memristors as synapses. Our focus is on how to use individual implement synaptic weight multiplications, in a way such that it not necessary (a) introduce global synchronization and (b) separate memristor learning phases from performing phases. the approaches described, neurons fire spikes asynchronously when they wish memristive synapses perform computation learn at their own pace,...
State-of-the-art image sensors suffer from significant limitations imposed by their very principle of operation. These acquire the visual information as a series "snapshot" images, recorded at discrete points in time. Visual gets time quantized predetermined frame rate which has no relation to dynamics present scene. Furthermore, each conveys all pixels, regardless whether this information, or part it, changed since last had been acquired. This acquisition method limits temporal resolution,...
This paper describes CAVIAR, a massively parallel hardware implementation of spike-based sensing-processing-learning-actuating system inspired by the physiology nervous system. CAVIAR uses asychronous address-event representation (AER) communication framework and was developed in context European Union funded project. It has four custom mixed-signal AER chips, five digital interface components, 45k neurons (spiking cells), up to 5M synapses, performs 12G synaptic operations per second,...
Event-driven visual sensors have attracted interest from a number of different research communities. They provide information in quite way conventional video systems consisting sequences still images rendered at given "frame rate." vision take inspiration biology. Each pixel sends out an event (spike) when it senses something meaningful is happening, without any notion frame. A special type event-driven sensor the so-called dynamic (DVS) where each computes relative changes light or...
Dynamic Vision Sensors (DVS) have recently appeared as a new paradigm for vision sensing and processing. They feature unique characteristics such contrast coding under wide illumination variation, micro-second latency response to fast stimuli, low output data rates (which greatly improves the efficiency of post-processing stages). can track extremely objects (e.g., time resolution is better than 100 kFrames/s video) without special lighting conditions. Their availability has triggered range...
Abstract Interdisciplinary research broadens the view of particular problems yielding fresh and possibly unexpected insights. This is case neuromorphic engineering where technology neuroscience cross-fertilize each other. For example, consider on one side recently discovered memristor, postulated in 1974, thanks to nanotechnology electronics. On other side, mechanism known as Spike-Time-Dependent-Plasticity (STDP) which describes a neuronal synaptic learning that outperforms traditional...
We show and validate a reliable circuit design technique based on source voltage shifting for current-mode signal processing down to femtoamperes. The involves specific-current extractors logarithmic current splitters obtaining on-chip subpicoampere currents. It also uses special sawtooth oscillator monitor measure currents few This way, are characterized without driving them off chip requiring expensive instrumentation with complicated low leakage setups. A mirror is introduced reliably...
Abstract A large effort is devoted to the research of new computing paradigms associated with innovative nanotechnologies that should complement and/or propose alternative solutions classical Von Neumann/CMOS (complementary metal oxide semiconductor) association. Among various propositions, spiking neural network (SNN) seems a valid candidate. i) In terms functions, SNN using relative spike timing for information coding are deemed be most effective at taking inspiration from brain allow fast...
This paper presents a 128 × dynamic vision sensor. Each pixel detects temporal changes in the local illumination. A minimum illumination contrast of 10% can be detected. compact preamplification stage has been introduced that allows to improve detectable over previous designs, while at same time reducing area by 1/3. The responds less than 3.6 μs. ability sensor capture very fast moving objects, rotating 10 K revolutions per second, verified experimentally. frame-based capable achieve this,...
This article reports on two databases for event-driven object recognition using a Dynamic Vision Sensor (DVS). The first, which we call Poker-DVS and is being released together with this article, was obtained by browsing specially made poker card decks in front of DVS camera 2-4 s. Each appeared the screen about 20-30 ms. pips were tracked isolated off-line to constitute 131-recording database. second database, MNIST-DVS December 2013, consists set 30,000 recordings displaying 10,000 moving...
This paper introduces a novel methodology for training an event-driven classifier within Spiking Neural Network (SNN) System capable of yielding good classification results when using both synthetic input data and real captured from Dynamic Vision Sensor (DVS) chips. The proposed supervised method uses the spiking activity provided by arbitrary topology prior SNN layers to build histograms train in frame domain stochastic gradient descent algorithm. In addition, this approach can cope with...
Event-Driven vision sensing is a new way of visual reality in frame-free manner. This is, the sensor (camera) not capturing sequence still frames, as conventional video and computer systems. In sensors each pixel autonomously asynchronously decides when to send its address out. way, output continuous stream events representing dynamically continuously without constraining frames. this paper we present an Convolution Module for computing 2D convolutions on such event streams. The has been...
We present a 32 times pixels contrast retina microchip that provides its output as an address event representation (AER) stream. Spatial is computed the ratio between pixel photocurrent and local average neighboring obtained with diffuser network. This current-based computation produces important amount of mismatch pixels, because currents can be low few pico-amperes. Consequently, compact calibration circuitry has been included to trimm each pixel. Measurements show reduction in standard...
This paper presents a modular, scalable approach to assembling hierarchically structured neuromorphic Address Event Representation (AER) systems. The method consists of arranging modules in 2D mesh, each communicating bidirectionally with all four neighbors. events include module label. Each includes an AER router which decides how route address events. Two routing approaches have been proposed, analyzed and tested, using either destination or source labels. Our analyses reveal that...
<para xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> We present a neuromorphic cortical-layer processing microchip for address event representation (AER) spike-based systems. The computes 2-D convolutions of video information represented in AER format real time. AER, as opposed to conventional frame-based representation, describes visual sequence events or spikes way similar biological brains. This allows fast identification and processing, without...
Neuromorphic circuits and systems techniques have great potential for exploiting novel nanotechnology devices, which suffer from parametric spread high defect rate. In this paper we explore some ways of building neural network sophisticated pattern recognition tasks using memristors. We will focus on spiking signal coding because its energy information efficiency, concentrate Convolutional Neural Networks their good scaling behavior, both in terms number synapses temporal processing delay....
Recent research in nanotechnology has led to the practical realization of nanoscale devices that behave as memristors, a device was postulated seventies by Chua based on circuit theoretical reasonings. On other hand, neuromorphic engineering, discipline implements physical artifacts neuroscience knowledge, related neural learning mechanisms operation memristors. As result, neuro-inspired architectures can be proposed exploit memristors for building very large scale systems with dense...
Most scene segmentation and categorization architectures for the extraction of features in images patches make exhaustive use 2D convolution operations template matching, search, denoising. Convolutional Neural Networks (ConvNets) are one example such that can implement general-purpose bio-inspired vision systems. In standard digital computers convolutions usually expensive terms resource consumption impose severe limitations efficient real-time applications. Nevertheless, neuro-cortex...
In computational neuroscience, synaptic plasticity learning rules are typically studied using the full 64-bit floating point precision computers provide. However, for dedicated hardware implementations, used not only penalizes directly required memory resources, but also computing, communication, and energy resources. When it comes to engineering, a key question is always find minimum number of necessary bits keep neurocomputational system working satisfactorily. Here we present some...
Object tracking is a major problem for many computer vision applications, but it continues to be computationally expensive. The use of bio-inspired neuromorphic event-driven dynamic sensors (DVSs) has heralded new methods processing, exploiting reduced amount data and very precise timing resolutions. Previous studies have shown these neural spiking well suited implementing single-sensor object systems, although they experience difficulties when solving ambiguities caused by occlusion. DVSs...