Setting specifications for the electronic implementation of biological neural-network-like vision systems on-chip is not straightforward, neither it is to simulate the resulting circuit. The structure of these systems leads to a netlist of more than 100.000 nodes for a small array of 100x150 pixels. Moreover, introducing an optical input in the low level simulation is nowadays not feasible with standard electrical simulation environments. Given that, to accomplish the task of integrating those systems in silicon to build compact, low power consuming, and reliable systems, a previous step in the standard analog electronic design flux should be introduced. Here a methodology to make the translation from the biological model to circuit-level specifications for electronic design is proposed. The purpose is to include non ideal effects as mismatching, noise, leakages, supply degradation, feedthrough, and temperature of operation in a high level description of the implementation, in order to accomplish behavioural simulations that require less computational effort and resources. A particular case study is presented, the analog electronic implementation of the locust’s Lobula Giant Movement Detector (LGMD), a neural structure that fires a collision alarm based on visual information. The final goal is a collision threat detection vision system on-chip for automotive applications.

This paper describes the architecture and retino-topic unit of a bio-inspired vision chip intended for automotive applications. The chip contains an array of 100X150 sensors which are able to capture high dynamic range (HDR) images, with a programmable compressive characteristic. The chip also incorporates a mechanism for adaptation of the global exposition time to the average illumination conditions. Average values are evaluated over image areas which are programmable by the user. In addition to the HDR pixel, every retino-topic unit in the array incorporates digital memory for three 6-bit pixel values (18-bits), as required for the implementation of a bionspired computing model for collisions detection which has been developed in the framework of a multidisciplinary European research project. All processing steps are executed off-chip, though we are currently working in the design of tiny digital processors (one per column) which will allow for running the whole model on-chip in a future version of this prototype. The chip has been designed in a 0.35μm 2P-4M technology and maintains its correct operation in extreme temperature conditions (from -40°C to 110°C).

In this paper a novel design and implementation of a VLSI Analogue Neural Net based on Multi-Layer Perceptron (MLP) with on-chip Back Propagation (BP) learning algorithm suitable for the resolution of classification problems is described. In order to implement a general and programmable analogue architecture, the design has been carried out in a hierarchical way. In this way the net has been divided in synapsis-blocks and neuron-blocks providing an easy method for the analysis. These blocks basically consist on simple cells, which are mainly, the activation functions (NAF), derivatives (DNAF), multipliers and weight update circuits. The analogue design is based on current-mode translinear techniques using MOS transistors working in the weak inversion region in order to reduce both the voltage supply and the power consumption. Moreover, with the purpose of minimizing the noise, offset and distortion of even order, the topologies are fully-differential and balanced. The circuit, named ANNE (Analogue Neural NEt), has been prototyped and characterized as a proof of concept on CMOS AMI-0.5A technology occupying a total area of 2.7mm2. The chip includes two versions of neural nets with on-chip BP learning algorithm, which are respectively a 2-1 and a 2-2-1 implementations. The proposed nets have been experimentally tested using supply voltages from 2.5V to 1.8V, which is suitable for single cell lithium-ion battery supply applications. Experimental results of both implementations included in ANNE exhibit a good performance on solving classification problems. These results have been compared with other proposed Analogue VLSI implementations of Neural Nets published in the literature demonstrating that our proposal is very efficient in terms of occupied area and power consumption.

In this paper we present our work analysing electroencephalographic (EEG) signals for the detection of seizure precursors in epilepsy. Volterra-systems and Cellular Nonlinear Networks are considered for a multidimensional signal analysis which is called the feature extraction problem throughout this contribution. Recent results obtained by applying a pattern detection algorithm and a nonlinear prediction of brain electrical activity will be discussed in detail. The aim of this interdisciplinary project is the realization of an implantable seizure warning and preventing system.

A random Boolean network is a synchronous Boolean automaton with n vertices. The parameters of an RBN can be tuned so that its statistical features match the characteristics of the gene regulatory system. The number of vertices of the RBN represents the number of genes in the cell. The number of cycles in the RBN's state space, called attractors, corresponds the number of different cell types. Attractor's length corresponds to the cell cycle time. Sensitivity of the attractors to different kind of perturbations, modeled by changing the state of a particular vertex, associated Boolean function, or network edge, reflects the stability of the cell to damage, mutations and virus attacks. In order to evaluate the attractors, their number and length have to be re-computed repeatedly. For large RBN's, searching for attractors in the O(2ⁿ) state space is an infeasible task. Fortunately, only a subset of vertices of an RBN, called relevant vertices, determines its dynamics. The remaining vertices are redundant. In this paper, we present an algorithm for identifying redundant vertices in RBNs which allows us to reduce the search space for computing attractors from O(2 ⁿ) to Θ2^√n. We also show how RBNs can be used for studying evolution.

For many epilepsy patients seizures cannot sufficiently be controlled by an antiepileptic pharmacatherapy. Furthermore, only in small number of cases a surgical treatment may be possible. The aim of this work is to contribute to the realization of an implantable seizure warning device. By using recordings of electroenzephalographical(EEG) signals obtained from the department of epileptology of the University of Bonn we studied a recently proposed algorithm for the detection of parameter changes in nonlinear systems. Firstly, after calculating the crosscorrelation function between the signals of two electrodes near the epileptic focus, a wavelet-analysis follows using a sliding window with the so called Mexican-Hat wavelet. Then the Shannon-Entropy of the wavelet-transformed data has been determined providing the information content on a time scale in subject to the dilation of the wavelet-transformation. It shows distinct changes at the seizure onset for all dilations and for all patients.

Efficient sparse coding of overcomplete transforms remains still an open problem. Different methods have been proposed in the literature, but most of them are limited by a heavy computational cost and by difficulties to find the optimal solutions. We propose here an algorithm suitable for Gabor wavelets and based on biological models. It is composed by local operations between neighboring transform coefficients and achieves a sparse representation with a relatively low computational cost. Used with a chain coder, this sparse Gabor wavelet transform is suitable for image compression but is also of interest also for other applications, in particular for edge and contour extraction and image denoising.

The study of spatio-temporal patterns generation and processing in systems with high parallelism like biological neuronal networks gives birth to a new technology able to realize architectures with robust performance even in noisy environments. The behavioural properties of neural assemblies warrant an effective exchange and use of information in presence of high-level neuronal noise. Neuron population processing and self-organization have been reproduced by connecting several neuron through synaptic connections, which can be either electrical or chemical, in artificial information processing architectures based on Field Programmable Gate Arrays (FPGA). The adopted neuron model is based on Izhikevich’s description of cortical neuron dynamics [1]. The development of biological neuronal network models has been focused on architecture features like changes over time of topologies, uniformity of the connections, node diversity, etc. The hardware reproduction of neuron dynamical behaviour, by giving high computation performance, allows the development of innovative computational methods and models based on self-organizing nonlinear architectures.

In this paper a biologically-inspired network of spiking neurons is used for robot navigation control. The implemented scheme is able to process information coming from the robot contact sensors in order to avoid obstacles and on the basis of these actions to learn how to respond to stimuli coming from range finder sensors. The implemented network is therefore able of reinforcement learning through a mechanism based on operant conditioning. This learning takes place according to a plasticity law in the synapses, based on spike timing. Simulation results discussed in the paper show the suitability of the approach and interesting adaptive properties of the network.

Address Event Representation (AER) is an emergent neuromorphic interchip communication protocol that allows for real-time virtual massive connectivity between huge number neurons located on different chips. By exploiting high speed digital communication circuits (with nano-seconds timings), synaptic neural connections can be time multiplexed, while neural activity signals (with mili-seconds timings) are sampled at low frequencies. Also, neurons generate 'events' according to their activity levels. More active neurons generate more events per unit time, and access the interchip communication channel more frequently, while neurons with low activity consume less communication bandwidth. When building multi-chip muti-layered AER systems it is absolutely necessary to have a computer interface that allows (a) to read AER interchip traffic into the computer and visualize it on screen, and (b) convert conventional frame-based video stream in the computer into AER and inject it at some point of the AER structure. This is necessary for test and debugging of complex AER systems. This paper addresses the problem of converting, in a computer, a conventional frame-based video stream into the spike event based representation AER. There exist several proposed software methods for synthetic generation of AER for bio-inspired systems. This paper presents a hardware implementation for one method, which is based on Linear-Feedback-Shift-Register (LFSR) pseudo-random number generation. The sequence of events generated by this hardware, which follows a Poisson distribution like a biological neuron, has been reconstructed using two AER integrator cells. The error of reconstruction for a set of images that produces different traffic loads of event in the AER bus is used as evaluation criteria. A VHDL description of the method, that includes the Xilinx PCI Core, has been implemented and tested using a general purpose PCI-AER board. This PCI-AER board has been developed by authors, and uses a Spartan II 200 FPGA. This system for AER Synthetic Generation is capable of transforming frames of 64x64 pixels, received through a standard computer PCI bus, at a frame rate of 25 frames per second, producing spike events at a peak rate of 10⁷ events per second.

Address Event Representation (AER) is an emergent neuromorphic interchip communication protocol that allows for real-time virtual massive connectivity between huge number neurons located on different chips. By exploiting high speed digital communication circuits (with nano-seconds timings), synaptic neural connections can be time multiplexed, while neural activity signals (with mili-seconds timings) are sampled at low frequencies. Also, neurons generate 'events' according to their activity levels. More active neurons generate more events per unit time, and access the interchip communication channel more frequently, while neurons with low activity consume less communication bandwidth. When building multi-chip muti-layered AER systems it is absolutely necessary to have a computer interface that allows (a) to read AER interchip traffic into the computer and visualize it on screen, and (b) inject a sequence of events at some point of the AER structure. This is necessary for testing and debugging complex AER systems. This paper presents a PCI to AER interface, that dispatches a sequence of events received from the PCI bus with embedded timing information to establish when each event will be delivered. A set of specialized states machines has been introduced to recovery the possible time delays introduced by the asynchronous AER bus. On the input channel, the interface capture events assigning a timestamp and delivers them through the PCI bus to MATLAB applications. It has been implemented in real time hardware using VHDL and it has been tested in a PCI-AER board, developed by authors, that includes a Spartan II 200 FPGA. The demonstration hardware is currently capable to send and receive events at a peak rate of 8,3 Mev/sec, and a typical rate of 1 Mev/sec.

We present a new platform for the optical analysis of biomolecules based upon the polysaccharide chitosan. The versatile, stable, and compatible nature of chitosan makes it an ideal material for integrating biological materials in microfabricated systems. Chitosan’s pH-responsive solubility allows electrochemical deposition, while its chemical reactivity enables facile coupling of proteins, oligonucleotides, and other biomolecules by covalent bonds. This work demonstrates the spatially selective assembly of a fluorescent molecule on chitosan and its applicability to microscale optical transducers. We define multimode waveguides and fluidic channels on a Pyrex wafer using a single layer of SU-8. Our implementation of sidewall patterning of transparent electrodes (indium tin oxide) on SU-8 structures is demonstrated and can be highly beneficial to fluorescent signal transduction. In this optical configuration, normally incident excitation light illuminates a chitosan surface on the vertical face of a collector waveguide intersected by a microfluidic channel. We demonstrate the collection of the optical signal in the integrated waveguide and analyze the signal by coupling the waveguide to a grating spectrometer.

This paper describes the fabrication, packaging and characterization of novel multilayer polymer microfluidic systems fabricated by a CMOS compatible process. These microfluidic devices were specially designed for BioMOEMS applications. Embedded multilayer rectangular smooth and uniform microchannels, 50 to 150 mm wide and 18mm deep were studied. Steady-state flow rates and pressure driven flow control were measured in the laminar flow regime. Flow rates ranging from 1 to 100 μl/min, at pressure drop ranging from 10 to 600 kPa, were obtained. These flow rates yield Reynolds numbers (Re) up to 20. Results indicate that the experimental Re and the flow friction coefficient (f) are in good agreement with the laminar flow theory. These experimental results facilitate the future designs of different microfluidic devices designed by using classical fluidic theory. We also present two different methods developed for macro/microfluidic packaging in order to connect these microfluidic devices to the macroscopic world. The microsystem packaging can withstand pressure drops up from 500 to 2000 kPa with any liquid leakage.

In this paper we examine the motion and behavior of particles suspended in aqueous solutions subjected to non-uniform ac electric fields. The particles can move due to forces exerted over them, or due to the motion of the surrounding liquid. In the first case, we have dielectrophoresis, due the action of ac electric fields over polarizable particles. The high strength electric fields often used in separation systems can give rise to fluid motion, which in turn results in a viscous drag on the particle. The electric field generates heat, leading to volume forces in the liquid. Gradients in conductivity and permittivity give rise to electrothermal forces; gradients in mass density to buoyancy. In addition, non-uniform ac electric fields produce forces on the induced charges in the diffuse double layer on the electrodes. This gives a steady fluid motion known as ac electroosmosis. We also discuss the effects of Brownian motion in this context. We calculate the different forces and displacements and compare them for a simple system consisting of a saline solution subjected to a traveling wave electric field. This example provides scaling laws of a wider applicability.

Address Event Representation (AER) is an emergent neuromorphic interchip communication protocol that allows for real-time virtual massive connectivity between huge number neurons located on different chips. By exploiting high speed digital communication circuits (with nano-seconds timings), synaptic neural connections can be time multiplexed, while neural activity signals (with mili-seconds timings) are sampled at low frequencies. Also, neurons generate 'events' according to their activity levels. More active neurons generate more events per unit time, and access the interchip communication channel more frequently, while neurons with low activity consume less communication bandwidth. In a typical AER transmitter chip, there is an array of neurons that generate events. They send events to a peripheral circuitry (let's call it "AER Generator") that transforms those events to neurons coordinates (addresses) which are put sequentially on an interchip high speed digital bus. This bus includes a parallel multi-bit address word plus a Rqst (request) and Ack (acknowledge) handshaking signals for asynchronous data exchange. There have been two main approaches published in the literature for implementing such "AER Generator" circuits. They differ on the way of handling event collisions coming from the array of neurons. One approach is based on detecting and discarding collisions, while the other incorporates arbitration for sequencing colliding events . The first approach is supposed to be simpler and faster, while the second is able to handle much higher event traffic. In this article we will concentrate on the second arbiter-based approach. Boahen has been publishing several techniques for implementing and improving the arbiter based approach. Originally, he proposed an arbitration squeme by rows, followed by a column arbitration. In this scheme, while one neuron was selected by the arbiters to transmit his event out of the chip, the rest of neurons in the array were freezed to transmit any further events during this time window. This limited the maximum transmission speed. In order to improve this speed, Boahen proposed an improved 'burst mode' scheme. In this scheme after the row arbitration, a complete row of events is pipelined out of the array and arbitered out of the chip at higher speed. During this single row event arbitration, the array is free to generate new events and communicate to the row arbiter, in a pipelined mode. This scheme significantly improves maximum event transmission speed, specially for high traffic situations were speed is more critical. We have analyzed and studied this approach and have detected some shortcomings in the circuits reported by Boahen, which may render some false situations under some statistical conditions. The present paper proposes some improvements to overcome such situations. The improved "AER Generator" has been implemented in an AER transmitter system

Address Event Representation (AER) is an emergent neuromorphic interchip communication protocol that allows for real-time virtual massive connectivity between huge number of neurons located on different chips. By exploiting high speed digital communication circuits (with nano-seconds timings), synaptic neural connections can be time multiplexed, while neural activity signals (with mili-seconds timings) are sampled at low frequencies. Also, neurons generate events according to their information levels. Neurons with more information (activity, derivative of activities, contrast, motion, edges,...) generate more events per unit time, and access the interchip communication channel more frequently, while neurons with low activity consume less communication bandwidth. AER technology has been used and reported for the implementation of various type of image sensors or retinae: luminance with local agc, contrast retinae, motion retinae,... Also, there has been a proposal for realizing programmable kernel image convolution chips. Such convolution chips would contain an array of pixels that perform weighted addition of events. Once a pixel has added sufficient event contributions to reach a fixed threshold, the pixel fires an event, which is then routed out of the chip for further processing. Such convolution chips have been proposed to be implemented using pulsed current mode mixed analog and digital circuit techniques. In this paper we present a fully digital pixel implementation to perform the weighted additions and fire the events. This way, for a given technology, there is a fully digital implementation reference against which compare the mixed signal implementations. We have designed, implemented and tested a fully digital AER convolution pixel. This pixel will be used to implement a full AER convolution chip for programmable kernel image convolution processing.

This paper reports a new model for electrical simulation of photodetector cells, that includes its complete dynamics, and enables full system characterization, both optical and electrical parts by using the same simulation environment (Spectre in our case). The modelling of the optical parts presented in this work allows the designer to change parameters such as incident spot position and optical power, speed in spot position, photodevices responsivity, pixel fill-factor, etc. The paper presents the design and the simulation-based verification of a Position Sensing Detection (PSD) system for applications with resolutions in the micrometer range and with spot movement tracking operation originated in a DNA sensing process.

New wireless technologies make possible the implementation of high level integration wireless devices which allow the replacement of traditional large wired monitoring devices. This kind of devices favours at-home hospitalization, reducing the affluence to sanitary assistance centers to make routine controls. This fact causes a really favourable social impact, especially for elder people, rural-zone inhabitant, chronic patients and handicapped people. Furthermore, it offers new functionalities to physicians and will reduce the sanitary cost. Among these functionalities, biomedical signals can be sent to other devices (screen, PDA, PC...) or processing centers, without restricting the patients' mobility. The aim of this project is the development and implementation of a reduced size multi-channel electrocardiograph based on BlueTooth, which allows wireless monitoring of patients, and the insertion of the information into the TCP/IP Hospital network.

The recent progress in both neurobiology and microelectronics suggests the creation of new, powerful tools to investigate the basic mechanisms of brain functionality. In particular, a lot of efforts are spent by scientific community to define new frameworks devoted to the analysis of in-vitro cultured neurons. One possible approach is recording their spiking activity to monitor the coordinated cellular behaviour and get insights about neural plasticity. Due to the nature of neurons action-potentials, when considering the design of an integrated microelectronic-based recording system, a number of problems arise. First, one would desire to have a high number of recording sites (i.e. several hundreds): this poses constraints on silicon area and power consumption. In this regard, our aim is to integrate-through on-chip post-processing techniques-hundreds of bio-compatible microsensors together with CMOS standard-process low-power (i.e. some tenths of uW per channel) conditioning electronics. Each recording channel is provided with sampling electronics to insure synchronous recording so that, for example, cross-correlation between signals coming from different sites can be performed. Extra-cellular potentials are in the range of [50-150] uV, so a comparison in terms of noise-efficiency was carried out among different architectures and very low-noise pre-amplification electronics (i.e. less than 5 uVrms) was designed. As spikes measurements are made with respect to the voltage of a reference electrode, we opted for an AC-coupled differential-input preamplifier provided with band-pass filtering capability. To achieve this, we implemented large time-constant (up to seconds) integrated components in the preamp feedback path. Thus, we got rid also of random slow-drifting DC-offsets and common mode signals. The paper will present our achievements in the design and implementation of a fully integrated bio-abio interface to record neural spiking activity. In particular, preliminary results will be reported.

In previous publications it has been shown that the prediction algorithm for multi-layer delay-type DTCNN may be used for the analysis of EEG-signals in order to find precursors of impending epileptic seizures. It has been stated that the application of time efficient training algorithms together with the consideration of symmetric templates lead to a significant decrease of the calculation complexity, allowing the analysis of long-term recordings of EEG-signals. In this contribution EEG-data, covering a total time of 6 days, were studied, applying the BFGS (Broiden-Fletcher-Goldfarb-Shanno) training method. To accomplish a very effective procedure, several symmetries have been tested and template structures leading to higher processing speed and optimal results have been implemented for the long-term studies. Distinct changes occuring before the onsets of impending seizures in the used data set were observed for different prediction parameters.

Short-Time Fourier transforms, Wigner-Ville distribution, and Wavelet Transforms have been commonly used when dealing with non-stationary signals, and they have been known as time-frequency distributions. Also, it is commonly intended to investigate the behaviour of phonocardiogram signals as a means of prediction some oh the pathologies of the human hart. For this, this paper aims to analyze the relationship between the instantaneous frequency of a PCG signal and the so-mentioned time-frequency distributions; three algorithms using Matlab functions have been developed: the first one, the estimation of the IF using the normalized linear moment, the second one, the estimation of the IF using the periodic first moment, and the third one, the computing of the WVD. Meanwhile, the computing of the STFT spectrogram is carried out with a Matlab function. Several simulations of the spectrogram for a set of PCG signals and the estimation of the IF are shown, and its relationship is validated through correlation. Finally, the second algorithm is a better choice because the estimation is not biased, whereas the WVD is very computing-demanding and offers no benefit since the estimation of the IF by using this TFD has an equivalent result when using the derivative of the phase of the analytic signal, which is also less computing-demanding.

Reaction-Diffusion systems can be applied to describe a broad class of nonlinear phenomena, in particular in biological systems and in the propagation of nonlinear waves in excitable media. Especially, pattern formation and chaotic behavior are observed in Reaction-Diffusion systems and can be analyzed. Due to their structure multi-layer Cellular Neural Networks (CNN) are capable of representing Reaction-Diffusion systems effectively. In this contribution Reaction-Diffusion CNN are considered for modeling dynamics of brain activity in epilepsy. Thereby the parameters of Reaction-Diffusion systems are determined in a supervised optimization process, and brain electrical activity using invasive multi-electrode EEG recordings is analyzed with the aim to detect of precursors of impending epileptic seizures. A detailed discussion of first results and potentiality of the proposed approach will be given.

In this paper we present an integrated ECG acquisition system implemented in a standard 0.35 μm CMOS technology. The system mainly consists of amplifiers, filters and an automatic offset compensation module. Composed of comparator, control logic and D/A converter, the compensator is an analog-digital mixed system and designed through a top-down approach. It is an independent module and can be easily reused. A low-pass filter of low cutoff frequency has been realized by using a Gm-C structure with very low bias current. Experimental measurements on the fabricated chip have been done. The results show that the integrated circuit works well with a low noise, a low consumption and a high efficacy of the offset compensator.

In this paper a bioinspired algorithm for collision detection is proposed, based on previous models of the locust (Locusta migratoria) visual system reported by F.C. Rind and her group, in the University of Newcastle-upon-Tyne. The algorithm is suitable for VLSI implementation in standard CMOS technologies as a system-on-chip for automotive applications. The working principle of the algorithm is to process a video stream that represents the current scenario, and to fire an alarm whenever an object approaches on a collision course. Moreover, it establishes a scale of warning states, from no danger to collision alarm, depending on the activity detected in the current scenario. In the worst case, the minimum time before collision at which the model fires the collision alarm is 40 msec (1 frame before, at 25 frames per second). Since the average time to successfully fire an airbag system is 2 msec, even in the worst case, this algorithm would be very helpful to more efficiently arm the airbag system, or even take some kind of collision avoidance countermeasures. Furthermore, two additional modules have been included: a "Topological Feature Estimator" and an "Attention Focusing Algorithm". The former takes into account the shape of the approaching object to decide whether it is a person, a road line or a car. This helps to take more adequate countermeasures and to filter false alarms. The latter centres the processing power into the most active zones of the input frame, thus saving memory and processing time resources.

In this contribution new results in the field of video processing regarding the problem of obstacle detection will be presented. Video sequences obtained from a camera mounted in a driving car are used as the input to a CNN and different templates are applied to extract multiple features from video sequences. Thereby, CNN with nonlinear weight functions have been considered allowing a reliable feature extraction. A detailed discussion of the algorithms and obtained results will be given in this paper.

A novel smart-sensor architecture is proposed, capable to segment and recognize characters in a monochrome image. It is capable to provide a list of ASCII codes representing the recognized characters from the monochrome visual field. It can operate as a blind's aid or for industrial applications. A bio-inspired cellular model with simple linear neurons was found the best to perform the nontrivial task of cropping isolated compact objects such as handwritten digits or characters. By attaching a simple outer-totalistic cell to each pixel sensor, emergent computation in the resulting cellular automata lattice provides a straightforward and compact solution to the otherwise computationally intensive problem of character segmentation. A simple and robust recognition algorithm is built in a compact sequential controller accessing the array of cells so that the integrated device can provide directly a list of codes of the recognized characters. Preliminary simulation tests indicate good performance and robustness to various distortions of the visual field.

A wide variety of image fusion techniques exist. A key term that is common to most is the "decision map". This map determines which information to take and at what place. Multifocus fusion deals with a stack of images that were acquired with a different focus point. In this case, one can say that the task of the decision map is to label parts that are in focus. If the focus length for each image in the stack is known, the decision map determines also a depth map that can be used for 3D surface reconstruction. Accuracy of the decision map is critical not only for image fusion itself, but even more for the surface reconstruction. Erroneous decisions can produce unrealistic glitches. We propose here to use information about image edges for increasing the accuracy of the decision map and enhancing in this way a standard wavelet-based fusion approach. We demonstrate the performance on real multifocus data under different noise levels.

Electrochemical oxidation and reduction taking place in films of a conducting polymer involve: generation and annihilation of positive charges on the polymeric chains, conformational changes along the polymeric chains, coulombic repulsions and generation of free volume with interchange of ions and water molecules between the polymer and the solution. So, electric pulses, conformational changes, ionic and aqueous interchanges are involved, as it occurs during most of the biological functions. Those changes induce, simultaneously, different electrochemical properties: electrochemomechanical by swelling and shrinking processes, electrochromic by change of the molecular orbitals, charge storage by accumulation of positive or negative charges, electron-ion transduction between an electronic conductor and an electrolyte. All those properties mimic biological functions: muscles, mimicking skins, electric organs or nervous pulses. Some of the developed devices as sensing actuators (muscles), or smart membranes are presented.

The interest in tactile sensors is increasing as their use in complex unstructured environments is demanded, like in telepresence, minimal invasive surgery, robotics etc. The matrix of pressure data these devices provide can be managed with many image processing algorithms to extract the required information. However, as in the case of vision chips or artificial retinas, problems arise when the array size and the computation complexity increase. Having a look to the skin, the information collected by every mechanoreceptor is not carried to the brain for its processing, but some complex pre-processing is performed to fit the limited throughput of the nervous system. This is specially important for high bandwidth demanding tasks. Experimental works report that neural response of skin mechanoreceptors encodes the change in local shape from an offset level rather than the absolute force or pressure distributions. This is also the behavior of the retina, which implements a spatio-temporal averaging. We propose the same strategy in tactile preprocessing, and we show preliminary results when it faces the detection of the slip, which involves fast real-time processing.

The precise control and manipulation of small masses of liquids is an important requirement in the lab-on-a-chip technology. Net fluid flows induced by ac potentials applied to arrays of co-planar interdigitated microelectrodes are reported. Two types of microelectrode structures have been studied: arrays of unequal width electrodes subjected to a single ac signal, and arrays of identical electrodes subjected to a travelling-wave potential. Experiments were performed using solutions of KCl in water of conductivities around 1mS/m placed on top of the electrodes. Fluorescent latex particles were used as tracers. In both microstructures, two fluid flow regimes have been observed: at small voltage amplitudes the fluid moves in a certain direction, and at higher voltage amplitudes the fluid flow is reversed. The fluid flow seems to be driven at the level of the electrodes in the two regimes. A theoretical model of ac electroosmosis is described. The model is based upon the Gouy-Chapman-Stern theory of the double layer. The theoretical results are in qualitative accordance with the experimental observations at low voltages.

Several microfluidic platforms incorporating cavities and channels have been designed and fabricated in silicon and fused silica. C4F8 and SF6 plasmas are used to etch reproducibly 400 μm features in silicon and 150 μm in fused silica. Hydrophilic surface characteristics allow capillary action without external pumping or electro-osmosis. Filling of poled cavities can be triggered by increasing temperature i.e. by tuning hydrophobicity of a channel. The pole structure can also be used for sieving particles of different size or elasticity. In this work, agarose beads trapped by poles were used for solid phase extraction. By covering the microfluidic features, filling is also achieved by cooling the substrate. Filling velocities of aqueous solutions have been observed to depend strongly on liquid composition, but also final treatment and roughness of silicon or silica surface. Mixing of two aqueous solutions can also be triggered by increasing temperature. Cavities with pre-immobilised substance can be filled simultaneously or, if necessary, sequentially. Various non-leaking 3D channel networks can be constructed by gluing, fusion or anodic bonding of many silicon or glass wafers including via holes. Integrating of electrical circuits for both silicon and silica is possible by standard IC technology.

This paper presents a neuron implementation based on floating-gate MOSFET (FG-MOS) structure. The computation is performed by charge distribution at the input of FG-MOS inverter determining the cell state. There is no current-flow through the interconnections after processing is completed, thus a significant reduction in DC power consumption can be achieved. Such neuron can be used to build a capacitively coupled cellular neural/nonlinear network (CNN) for processing black and white (B/W) images. Although the coupling coefficients are basically implemented with capacitances, this approach provides them with 1-bit programmability. Also the neuron's threshold level can be digitally programmed to four different values. The templates operating on the B/W images can be modified to have only binary-valued {0,1} terms or can be split into such (sequentially run) simple subtasks. Therefore, the presented neuron structure is able to perform the evaluation of almost all B/W templates proposed so far. Exploration of FG-MOS structures can help to understand the implementation problems of capacitively coupled CNNs. Such a situation appears, e.g., in nanoelectronic CNNs composed of single-electron tunneling (SET) transistors, which also deal with B/W images only. Moreover, the binary programmability approach utilized here should help to develop an effective programming scheme for future SET or CMOS-SET hybrid CNN implementations. Along with the neuron structure, its operation description and simulation results of the 8 x 8 network are presented.

This paper explores different alternatives to carry out a model of digital CNN from the point of view of its implementation on FPGAs. It shows the developments of four different DT-CNN models obtained from different transformations made to the original continuous model of CNN. Next, each discrete approach is simulated and compared with the rest of approaches and the continuous models. The objective of this study is to find the approach which best emulates the continuous neuron model at minimum computational cost. The simulations and temporal analysis of the discrete models have been made both in feedback and open system in order to verify their functionality. Finally, the architecture of the best model is implementated on an FPGA obtaining very interesting results.

Many studies in neuroscience have shown that nonlinear dynamic networks represent a bio-inspired models for information and image processing. Recent studies on the thalamo-cortical system have shown that weakly connected oscillatory networks have the capability of modelling the architecture of a neurocomputer. In particular they have associative properties and can be exploited for dynamic pattern recognition. In this manuscript the global dynamic behavior of weakly connected cellular networks of oscillators are investigated. It is assumed that each cell admits of a Lur'e description. In case of weak coupling the main dynamic features of the network are revealed by the phase deviation equation (i.e. the equation that describes the phase deviation due to the weak coupling). Firstly a very accurate analytic expression of the phase deviation equation is derived via the joint application of the describing function technique and of Malkin's Theorem. Then it is shown that the total number of periodic limit cycles with their stability properties can be estimated through the analysis of the phase deviation equation.

The paper investigates the potential for a packet switching network for real-time image processing by a Cellular Neural Network (CNN) implemented on a Field-Programmable Gate-Array (FPGA). The implementation of a CNN requires several parameter restrictions with respect to the universal concept. For instance, the number representation and the cloning template are often confined to respectively 8 bits and a neighborhood of 1. It has been shown that optimal (i.e. minimal level) CNN architectures as derived from a morphological specification of the desired operation lead to arbitrarily large templates. A subsequent transformation step can turn this into a sequence of smaller templates for a specified hardware platform. The existence of a generic platform that can already handle the universal CNN architecture for prototyping and verification eliminates this need for technology-driven performance degradation. The proposed packet switcher consists of a physical layer where the CNN nodal function is performed, a data-link layer where the nodal data are maintained, a network layer with the packet receiver and sender and the actual switch as element of the transport layer. This ISO/OSI compliant level-wise structure monitors the network parameters and autonomously adjusts for the size of the neighborhood. It separates the broadcast of the network variables from the actual computation, allowing each to be executed at its own speed. The concept is tested on a re-design of the ILVA architecture and has been shown to handle arbitrary neighborhoods and precision at a comparable size and speed (1 node per BlockRAM / multiplier module @220 MHz clock).

Recently CNN with nonlinear weight functions are used for various problems. Thereby nonlinear weights are represented by polynomials or tabulated functions combined with a cubic spline interpolation. In this paper a linear interpolation technique is considered to allow an accurate approximation of nonlinear weight functions in CNN. In a previous publication the Table Minimising Algorithm (TMA) was introduced and applied to the Korteweg-de Vries-equation (KdV). In this contribution new results obtained by applying the algorithm to additional partial differential equations (PDE) will be given and discussed.

In this paper a model for auditory perception is introduced. This model is based on a network of integrate-and-fire and resonate-and-fire neurons and is aimed to control the phonotaxis behavior of a roving robot. The starting point is the model of phonotaxis in Gryllus Bimaculatus: the model consists of four integrate-and-fire neurons and is able of discriminating the calling song of male cricket and orienting the robot towards the sound source. This paper aims to extend the model to include an amplitude-frequency clustering. The proposed spiking network shows different behaviors associated with different characteristics of the input signals (amplitude and frequency). The behavior implemented on the robot is similar to the cricket behavior, where some frequencies are associated with the calling song of male crickets, while other ones indicate the presence of predators. Therefore, the whole model for auditory perception is devoted to control different responses (attractive or repulsive) depending on the input characteristics. The performance of the control system has been evaluated with several experiments carried out on a roving robot.

In this paper we propose a compact current mode calibration technique for analog current mode circuits operating in weak inversion. The technique is based on the use of MOS ladder structures controlled by a locally stored digital calibration word. Extensive Monte Carlo have been performed to analyze the technique. The circuit has been applied to calibrate a circuit intended for use in a contrast retina.

Phylogenetic analysis is a branch of biology that establishes the evolutionary relationships between living organisms. The goal of phylogenetic analysis is to determine the order and approximate timing of speciation events in the evolution of a given set of species. Phylogenetic networks allow to represent evolutionary histories that include events like recombination and hybridization. In this paper, we introduce a class of phylogenetic networks called extended galled-trees in which recombination cycles share no edge. We show that the site consistency problem, which is NP-hard in general, can be solved in polynomial time for this class of phylogenetic networks.

An FPGA-based approach is proposed to build an augmented reality system in order to aid people affected by a visual disorder known as tunnel vision. The aim is to increase the user's knowledge of his environment by superimposing on his own view useful information obtained with image processing. Two different alternatives have been explored to perform the required image processing: a specific purpose algorithm to extract edge detection information, and a cellular neural network with the suitable template. Their implementations in reconfigurable hardware pursue to take advantage of the performance and flexibility that show modern FPGAs. This paper describes the hardware implementation of both the Canny algorithm and the cellular neural network, and the overall system architecture. Results of the implementations and examples of the system functionality are presented.

The auscultation of the heart is still the first basic analysis tool used to evaluate the functional state of the heart, as well as the first indicator used to submit the patient to a cardiologist. In order to improve the diagnosis capabilities of auscultation, signal processing algorithms are currently being developed to assist the physician at primary care centers for adult and pediatric population. A basic task for the diagnosis from the phonocardiogram is to detect the events (main and additional sounds, murmurs and clicks) present in the cardiac cycle. This is usually made by applying a threshold and detecting the events that are bigger than the threshold. However, this method usually does not allow the detection of the main sounds when additional sounds and murmurs exist, or it may join several events into a unique one. In this paper we present a reliable method to detect the events present in the phonocardiogram, even in the presence of heart murmurs or additional sounds. The method detects relative maxima peaks in the amplitude envelope of the phonocardiogram, and computes a set of parameters associated with each event. Finally, a set of characteristics is extracted from each event to aid in the identification of the events. Besides, the morphology of the murmurs is also detected, which aids in the differentiation of different diseases that can occur in the same temporal localization. The algorithms have been applied to real normal heart sounds and murmurs, achieving satisfactory results.

Cellular patterning plays an essential role in the development of cell-based biosensors, cell culture analogues and tissue engineering. In particular addressability at cell level is needed for the recording of neuronal activity and interpretation of neuronal communication, thus providing insight in learning and memory processes as well as in the influence of drugs on neuronal activity. In this paper, we report an approach to guide neuronal growth into simplified networks with single cell resolution. This enables direct correlation of individual neurons with relevant positions on a chip surface and can improve activity recordings by means of microelectrodes1 or field-effect transistors. Our method is based on aligned micro contact printing and allows the deposition of different types of guidance cues with micrometer precision in an easy manner. Making use of a flipchip bonder, we have created patterns of poly-L-lysine (PLL) and laminin to guide the development of neuronal networks. To ensure long-term stability of a neuronal pattern, chemical guidance cues alone do not suffice. Cell compliance to the cytophilic pattern seldomly extends beyond the duration of one week in culture. We have therefore made use of the hydrophobic properties of fluoropolymers such as teflon to discourage neuronal adhesion and at the same time create topography to hold the neurons mechanically in place. This cytophobic pattern was complemented with an aligned deposition of PLL between the teflon lanes. In this paper, aligned micro contact printing, combined chemical and topographical functionalisation of surfaces and the results of the evaluation in primary hippocampal cultures by immunocytochemical staining will be discussed.

In first place, in this paper, the basic process of parallel code implementation is discussed for a VLIW architecture. Parallel code modules allow the implementation of a contour active (snake) for segmentation and tracking of endocardium in echocardiographic images. In second place, this work discusses the performance obtained through this design model. In this case, it is necessary to check performances in order to obtain a qualitative and quantitative measurement of our implementation. We have chosen one example which permits to understand the methodology used in order to obtain the maximum performance of hardware features of VLIW processor: the distance between points of active contour, very used in different modules composing the active contour algorithm.

Interdigital (IDT) microsensors are one of the most commonly used periodic microelectrode devices in a wide range of fields such as microelectromechanical systems (MEMS), telecommunications, chemical sensing, etc. IDT biochemical sensors targeted towards the direct detection of immobilized ssDNA (single strand nucleic acid sequences) and the subsequent hybridization with the complementary strand are currently an area of significant research interest. The most common outputs of measurement are changes in resistance and capacitance between electrodes as a result of changes in the conductivity or dielectric constant of a thin layer of material or solution coating the IDT device, which contains the oligomeric DNA of interest. DNA may be immobilized on both the electrode surfaces and the interdigital spaces or only on the latter, depending on the method used for the chemical modification of the sensor surface. In this work, various IDT designs are explored from the point of view of sensitivity to the changes in impedance associated with modifications in the conductivity of the material near the electrodes. The designs are studied by an electroquasistatics, finite element method-based 3D model to simulate the variation of IDT sensor impedance. A range of device geometries are considered, with particular attention paid to the width of the fingers relative to the period, or metallization ratio. Our results show that low metallization ratios lead to better impedance sensitivity. The simulation models have also been checked experimentally with commercially available IDTs, where a good agreement has been obtained between calculated and measured impedance.

Synaptic transmission in neuronal networks occur on a very short time scale and is highly specific. Fast, sensitive and in situ detection of single neuron L-glutamate release is essential for the investigation of these events under physiological or pathophysiological conditions. Up till now, amperometry with enzyme-modified electrodes has extensively been used to monitor extracellular glutamate release. However, due to in situ signal amplification, ENzyme-modified Field-Effect Transistors (ENFETs) have the advantage of preserving sensitivity and a fast response time when scaled down to micrometer dimensions. We have realized a L-GLutamate OxiDase (GLOD) functionalized FET to be used for glutamate detection in neuronal cultures. Effective and reproducible immobilization of GLOD on the FET active area is achieved by using Poly-L-Lysine (PLL) as a loading matrix. PLL plays a dual role in the assay: on the one hand this molecule serves as a platform for obtaining high enzyme loading and on the other hand it benefits the survival of the neuronal network on the active area of the FET. Both PLL and enzyme immobilization were characterised by quartz crystal microbalance measurements. A much higher enzyme loading has been achieved by this approach compared to immobilization methods without PLL. The enzyme coating has proven to be extremely durable as it keeps its activity for at least 3 weeks as monitored by a colorimetric assay. FET characterisation curves and glutamate response curves of the ENFET are presented.

This paper describes a new blind deconvolution method implemented by means of an evolutionary algorithm (EA), designed as a multi-objective optimization problem (MOP) approach. The last generation of the EA can be used for selecting the best image under a given quality criterion or submitted to a fusion method for producing an improved result. The fusion procedure is preferred here and implemented through a new robust method, based on the local space-frequency information extracted from the Wigner distribution of the image. This fusion method has been recently developed by the authors providing excellent experimental results.

The noise problem, such as the fixed pattern noise (FPN) due to the process variation, should be considered when designing a vision chip. In this paper, we proposed an edge detection circuit based on biological retina using an offset-free column readout circuit (OFCRC) to reduce the FPN occurring in the photo-detector. The OFCRC consists of one source follower, one capacitor and five transmission gates. Thus, it is simpler than a conventional correlated double sampling (CDS) circuit. A vision chip for edge detection has been designed and fabricated using a 0.35μm 2-poly 4-metal CMOS process and its output characteristics have been investigated.