A5081607982- Advanced Memory and Neural Computing
- Neuroscience and Neural Engineering
- CCD and CMOS Imaging Sensors
- Neural dynamics and brain function
- Advanced Vision and Imaging
- Image Processing Techniques and Applications
- Photoreceptor and optogenetics research
- Optical measurement and interference techniques
- Radiation Detection and Scintillator Technologies
- Radiation Effects in Electronics
- Advanced Neural Network Applications
- Neural Networks and Reservoir Computing
- Embedded Systems Design Techniques
- Neural Networks and Applications
Universidad de Sevilla
2006-2016
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,...
In this paper we present a neuro-inspired spike-based close-loop controller written in VHDL and implemented for FPGAs. This has been focused on controlling DC motor speed, but only using spikes information representation, processing driving. It could be applied to other motors with proper driver adaptation. architecture represents one of the latest layers Spiking Neural Network (SNN), which implements bridge between robotics actuators sensors. The presented control system fuses actuation...
Neuromorphic engineers study models and implementations of systems that mimic neurons behavior in the brain. Neuro-inspired commonly use spikes to represent information. This representation has several advantages: its robustness noise thanks repetition, continuous analog information using digital pulses, capacity pre-processing during transmission time, ..., Furthermore, is an efficient way, found by nature, codify, transmit process In this paper we propose, design, analyze neuro-inspired...
Image convolution operations in digital computer systems are usually very expensive terms of resource consumption (processor resources and processing time) for an efficient Real-Time application. In these scenarios the visual information is divided into frames each one has to be completely processed before next frame arrives order warranty real-time. A spike-based philosophy computing convolutions based on neuro-inspired Address-Event-Representation (AER) achieving high performances. this...
Address-event-representation (AER) is a communication protocol for transferring asynchronous events between VLSI chips, originally developed neuro-inspired processing systems (for example, image processing). Such may consist of complicated hierarchical structure with many chips that transmit data among them in real time, while performing some convolutions). The information transmitted sequence spikes coded using high speed digital buses. These multi-layer and multi-chip AER perform actually...
Address-Event-Representation (AER) is an asynchronous protocol for transferring the information of spiking neuro-inspired systems.Actually AER systems are able to see, ear, process information, and learn.Regarding actuation step, has been used implementing Central Pattern Generator algorithms, but not controlling actuators in a closed-loop spike-based way.In this paper we analyze based model real-time control system.We demonstrate it into differential system two-wheel vehicle using feedback...
In this paper, we explore the capabilities of a sound classification system that combines both novel FPGA cochlear model implementation and bio-inspired technique based on trained convolutional spiking network. The neuromorphic auditory is used in work produces form representation analogous to spike outputs biological cochlea. has been developed using set spike-based processing building blocks frequency domain. They band pass filters spike-domain splits audio information 128 channels, 64 for...
Address event representation (AER) is a neuromorphic interchip communication protocol that allows for real-time connectivity between huge number neurons located on different chips. By exploiting high speed digital circuits (nano-seconds), synaptic neural connections can be time multiplexed (mili-seconds). When building multi-chip muti-layered AER systems it absolutely necessary to have computer interface allows: (a) read traffic; and (b) inject sequence of events the structure. This paper...
Neuromorphic engineering tries to mimic biological information processing. Address-Event-Representation (AER) is an asynchronous protocol for transferring the of spiking neuro-inspired systems. Currently AER systems are able sense visual and auditory stimulus, process information, learn, control robots, etc. In this paper we present based layer correct in real time tilt vision sensor, using a high speed algorithmic mapping layer. A co-design platform (the AER-Robot platform), with Xilinx...
Neuromorphic engineering tries to mimic biology in information processing. Address-event representation (AER) is a neuromorphic communication protocol for spiking neurons between different layers. AER bio-inspired image sensor are called ldquoretinardquo. This kind of sensors measure visual not based on frames from real life and generates corresponding events. In this paper we provide an alternative, cheap FPGA, that takes images provided by analog video source (video composite signal),...
Address-event representation (AER) is a neuromorphic communication protocol for transferring information of spiking neurons implemented into VLSI chips. These neuro-inspired implementations have been used to design sensor chips (retina, cochleas), processing (convolutions, filters) and learning chips, what makes possible the development complex, multilayer, multichip systems. In biology one last steps move muscle, apply results these complex real world. One interesting question be able...
Image processing in digital computer systems usually considers the visual information as a sequence of frames. These frames are from cameras that capture reality for short period time. They renewed and transmitted at rate 25–30 fps (typical real-time scenario). Digital video has to process each frame order obtain filter result or detect feature on input. In stereo vision, existing algorithms use two them pixel by until it is found pattern match section both Spike-based relatively new...
Distance calculation is always one of the most important goals in a digital stereoscopic vision system. In an AER system this goal very too, but it cannot be calculated as accurately we would like. This demonstration shows first approximation field, using disparity algorithm between both retinas. The can make distance approach about moving object, more specifically, qualitative estimation. Taking into account stereo features, previous retina positioning and Hold&Fire building block, are able...
Neuromorphic engineering tries to mimic biological information processing. Address-event representation (AER) is a neuromorphic communication protocol for spiking neurons between different chips. We present new way drive robotic platforms using neurons. have simulated control models DC motors, and developed mobile robot (Eddie) controlled only by spikes. apply AER the control, monitoring measuring spike activity inside robot. The AER-Robot tool, sent PC USBAERmini2 interface.
Correcting digital images tilt needs huge quantities of memory, high computational resources, and use to take a considerable amount time. This demonstration shows how spikes-based silicon retina dynamic vision sensor (DVS) can corrected in real time using commercial accelerometer. DVS output is stream spikes codified the address-event representation (AER). Event-based processing focused on change addresses. Taking into account this feature, we present an AER based layer able correct tilt,...
A high speed sample image processing application using AER-based components is presented. The setup objective to distinguish between two propellers of different shape rotating at (around 1000 revolutions/sec) show event-based systems capabilities in applications. Event-based schemes allow the most relevant information propagate faster through system layers. So sped up because a rough result may be available when only little part input has arrived. This much than conventional frame-based they...