This PDF file contains the front matter associate with SPIE Porceedings Volume 7365, including the Title Page, Copyright information, Table of Contents, Introduction (if any), and the Conference Committee listing.

Stand-alone applications of vision are severely constrained by their limited power budget. This is one of the main reasons why vision has not yet been widely incorporated into wireless sensor networks. For them, image processing should be suscribed to the sensor node in order to reduce network traffic and its associated power consumption. In this scenario, operating the conventional acquisition-digitization-processing chain is unfeasible under tight power limitations. A bio-inspired scheme can be followed to meet the timing requirements while maintaining a low power consumption. In our approach, part of the low-level image processing is conveyed to the focal-plane thus speeding up system operation. Moreover, if a moderate accuracy is permissible, signal processing is realized in the analog domain, resulting in a highly efficient implementation. In this paper we propose a circuit to realize dynamic texture segmentation based on focal-plane spatial bandpass filtering of image subdivisions. By the appropriate binning, we introduce some constrains into the spatial extent of the targeted texture. By running time-controlled linear diffusion within each bin, a specific band of spatial frequencies can be highlighted. Measuring the average energy of the components in that band at each image bin the presence of a targeted texture can be detected and quantified. The resulting low-resolution representation of the scene can be then employed to track the texture along an image flow. An application specific chip, based on this analysis, is being developed for natural spaces monitoring by means of a network of low-power vision systems.

Connectivity in the human retina is complex. Over one hundred million photoreceptors transduce light into electrical signals. These electrical signals are sent to the ganglion cells through amacrine and bipolar cells. Lateral connections involving horizontal and amacrine cells span throughout the outer plexiform layer and inner plexiform layer respectively. Horizontal cells are important for photoreceptor regulation by depolarizing them after an illumination occurs. Horizontal cells themselves form an electrical network that communicates by gap junctions, and these cells exhibit plasticity (change in behavior and structure) with respect to glycine receptors. The bipolar and amacrine cells transfer electrical signals from photoreceptors to the ganglion cells. Furthermore, amacrine cells are responsible for further processing the retinal image. Finally, the ganglion cells receive electrical signals from the bipolar and amacrine cells and will spike at a faster rate if there is a change in the overall intensity for a group of photoreceptors, sending a signal to the brain. Dramatic progress is being made with respect to retinal prostheses, raising hope for an entire synthetic retina in the future. We propose a bio-inspired 3D hierarchical pyramidal architecture for a synthetic retina that mimics the overall structure of the human retina. We chose to use a 3D architecture to facilitate connectivity among retinal cells, maintaining a hierarchical structure similar to that of the biological retina. The first layer of the architecture contains electronic circuits that model photoreceptors and horizontal cells. The second layer contains amacrine and bipolar electronic cells, and the third layer contains ganglion cells. Layer I has the highest number of cells, and layer III has the lowest number of cells, resulting in a pyramidal architecture. In our proposed architecture we intend to use photodetectors to transduce light into electrical signals. We propose to employ wireless communication to mimic the gap junction behavior among horizontal cells. These cells could communicate laterally to neighboring horizontal cells through a network of spin wave transmitters and receivers that send magnetic waves over the surface of the first layer of the synthetic retina. We discuss the tradeoffs for having point-to-point connections versus a network on chip in the second layer. We examine the use of 3D CMOS technologies as well as nanotechnologies for the implementation of this retina, considering size, interconnectivity capabilities, and power consumption. Finally, we estimate the volume, delay and power dissipation of our architecture.

Detection of the visual scene by the eye and the resultant neural interactions of the retina-brain system give us our perception of sight. We have developed an Active Pixel Sensor (APS) to be used as a tool for both furthering understanding of these interactions via experimentation with the retina and to make developments towards a realisable retinal prosthesis. The sensor consists of 469 pixels in a hexagonal array. The pixels are interconnected by a programmable neural network to mimic lateral interactions between retinal cells. Outputs from the sensor are in the form of biphasic current pulse trains suitable to stimulate retinal cells via a biocompatible array. The APS will be described with initial characterisation and test results.

Segmentation is the process of representing a digital image into multiple meaningful regions. Since these applications require more computational power in real time applications, we have implemented a new segmentation algorithm using the capabilities of Eye-RIS Vision System to execute the algorithm in very short time. The segmentation algorithm is implemented mainly in three steps. In the first step, which is pre-processing step, the images are acquired and noise filtering through Gaussian function is performed. In the second step, Sobel operators based edge detection approach is implemented on the system. In the last step, morphologic and logic operations are used to segment the images as post processing. The experimental results performed for different images show the accuracy of the proposed segmentation algorithm. Visual inspection and timing analysis (7.83 ms, 127 frame/sec) prove that the proposed segmentation algorithm can be executed for real time video processing applications. Also, these results prove the capability of Eye-RIS Vision System for real time image processing applications

This paper describes a correlation-based navigation algorithm, based on an unsupervised learning paradigm for spiking neural networks, called Spike Timing Dependent Plasticity (STDP). This algorithm was implemented on a new bio-inspired hybrid mini-robot called TriBot to learn and increase its behavioral capabilities. In fact correlation based algorithms have been found to explain many basic behaviors in simple animals. The main interesting consequence of STDP is that the system is able to learn high-level sensor features, based on a set of basic reflexes, depending on some low-level sensor inputs. TriBot is composed of 3 modules, the first two being identical and inspired by the Whegs hybrid robot. The peculiar characteristics of the robot consists in the innovative shape of the three-spoke appendages that allow to increase stability of the structure. The last module is composed of two standard legs with 3 degrees of freedom each. Thanks to the cooperation among these modules, TriBot is able to face with irregular terrains overcoming potential deadlock situations, to climb high obstacles compared to its size and to manipulate objects. Robot experiments will be reported to demonstrate the potentiality and the effectiveness of the approach.

In this paper a new general purpose perceptual control architecture is presented and applied to robot navigation in cluttered environments. In nature, insects show the ability to react to certain stimuli with simple reflexes using direct sensory-motor pathways, which can be considered as basic behaviors, while high brain regions provide secondary pathway allowing the emergence of a cognitive behavior which modulates the basic abilities. Taking inspiration from this evidence, our architecture modulates, through a reinforcement learning, a set of competitive and concurrent basic behaviors in order to accomplish the task assigned through a reward function. The core of the architecture is constituted by the Representation layer, where different stimuli, triggering competitive reflexes, are fused to form a unique abstract picture of the environment. The representation is formalized by means of Reaction-Diffusion nonlinear partial differential equations, under the paradigm of the Cellular Neural Networks, whose dynamics converges to steady-state Turing patterns. A suitable unsupervised learning, introduced at the afferent (input) stage, leads to the shaping of the basins of attractions of the Turing patterns in order to incrementally drive the association between sensor stimuli and patterns. In this way, at the end of the leaning stage, each pattern is characteristic of a particular behavior modulation, while its trained basin of attraction contains the set of all the environment conditions, as recorded through the sensors, leading to the emergence of that particular behavior modulation. Robot simulations are reported to demonstrate the potentiality and the effectiveness of the approach.

This paper introduces a simple algorithm to solve robot path finding problem using active wave computing techniques. A two-dimensional Cellular Neural/Nonlinear Network (CNN), consist of relaxation oscillators, has been used to generate active waves and to process the visual information. The network, which has been implemented on a Field Programmable Gate Array (FPGA) chip, has the feature of being programmed, controlled and observed by a host computer. The arena of the robot is modelled as the medium of the active waves on the network. Active waves are employed to cover the whole medium with their own dynamics, by starting from an initial point. The proposed algorithm is achieved by observing the motion of the wave-front of the active waves. Host program first loads the arena model onto the active wave generator network and command to start the generation. Then periodically pulls the network image from the generator hardware to analyze evolution of the active waves. When the algorithm is completed, vectorial data image is generated. The path from any of the pixel on this image to the active wave generating pixel is drawn by the vectors on this image. The robot arena may be a complicated labyrinth or may have a simple geometry. But, the arena surface always must be flat. Our Autowave Generator CNN implementation which is settled on the Xilinx University Program Virtex-II Pro Development System is operated by a MATLAB program running on the host computer. As the active wave generator hardware has 16, 384 neurons, an arena with 128 × 128 pixels can be modeled and solved by the algorithm. The system also has a monitor and network image is depicted on the monitor simultaneously.

This paper aims to describe how the AnaFocus' Eye-RIS family of vision systems has been successfully embedded within the roving robots developed under the framework of SPARK and SPARK II European projects to solve the action-oriented perception problem in real time. Indeed, the Eye-RIS family is a set of vision systems which are conceived for single-chip integration using CMOS technologies. The Eye-RIS systems employ a bio-inspired architecture where image acquisition and processing are truly intermingled and the processing itself is carried out in two steps. At the first step, processing is fully parallel owing to the concourse of dedicated circuit structures which are integrated close to the sensors. These structures handle basically analog information. At the second step, processing is realized on digitally-coded information data by means of digital processors. On the other hand, SPARK I and SPARK II are European research projects which goal is to develop completely new sensing-perceiving-moving artefacts inspired by the basic principles of living systems and based on the concept of "selforganization". As a result, its low-power consumption together with its huge image-processing capabilities makes the Eye-RIS vision system a suitable choice to be embedded within the roving robots developed under the framework of SPARK projects and to implement in real time the resulting mathematical models for action-oriented perception.

For the detection of forest fires the "fire-loving" (pyrophilous) jewel beetle Melanophila acuminata uses a pair of sensor arrays each consisting of about 90 infrared (IR) receptors which are located on either sides of the body. The IR receptors most likely have evolved from common contact hair mechanoreceptors. Compared to a mechanoreceptor, an IR receptor shows the following special features: (i) the formation of a complex cuticular sphere instead of the bristle; the sphere consists of a hard outer exocuticular shell as well as of an inner softer and spongy mesocuticular core. (ii) The enclosure of the dendritic tip of the mechanosensitive neuron inside the sphere in a fluid-filled inner pressure chamber which is connected with the system of microcavities and nanocanals in the mesocuticular core. Hence we propose that an IR sensillum most probably acts as a microfluidic converter of infrared radiation into an increase in internal pressure inside the sphere which is measured by the mechanosensitive neuron. Because the miniaturized receptors respond within a few milliseconds to a brief pulse of IR radiation an approach is made to develop technical IR sensors based on the Melanophila IR receptors. Numerical simulations of sensor performance suggest that the sensitivity of a single IR receptor is in the range of 15 mW/cm². Theoretical calculations which are based on a hypothetical fire of defined temperature and size demonstrate that a beetle should be able to detect a forest fire from a distance of 10 km. A fluidfilled Golay cell was taken as a basis for the design of a first sensor prototype.

Instead of vision, many animals use alternative senses for object detection. Weakly electric fish employ "active electrolocation", during which they discharge an electric organ emitting electrical current pulses (electric organ discharges, EOD). Local EODs are sensed by electroreceptors in the fish's skin, which respond to changes of the signal caused by nearby objects. Fish can gain information about attributes of an object, such as size, shape, distance, and complex impedance. When close to the fish, each object projects an 'electric image' onto the fish's skin. In order to get information about an object, the fish has to analyze the object's electric image by sampling its voltage distribution with the electroreceptors. We now know a great deal about the mechanisms the fish use to gain information about objects in their environment. Inspired by the remarkable capabilities of weakly electric fish in detecting and recognizing objects with their electric sense, we are designing technical sensor systems that can solve similar sensing problems. We applied the principles of active electrolocation to devices that produce electrical current pulses in water and simultaneously sense local current densities. Depending on the specific task, sensors can be designed which detect an object, localize it in space, determine its distance, and measure certain object properties such as material properties, thickness, or material faults. We present first experiments and FEM simulations on the optimal sensor arrangement regarding the sensor requirements e. g. localization of objects or distance measurements. Different methods of the sensor read-out and signal processing are compared.

Many artificial skins for robotics are based on piezoresistive films that cover an array of electrodes. Local preprocessing is a must in these systems to reduce errors and interferences and cope with the large amount of data provided by the sensor. This paper presents circuitry based on an FPGA to implement the interface to the artificial skin. The approach consists of a direct connection. The analog to digital conversion procedure is simple. It consists of measuring the discharging time of a capacitor through the resistance we want to read. This first proposed approach needs isolated tactels, so the raw sensor has to be fabricated in this way. If the tactile array is large, the strategy is not feasible. For instance, up to 288 pins are required to implement the interface with an array of 16x16 tactels. The proposal of this work for this case is to replace passive integrators by active ones. The result is a circuitry that allows the cancellation of interferences due to parasitic resistors and the sharing of the addressing tracks. Moreover, the FPGA allows the processing of data from the tactile sensor at a very high rate. This is because the high number of I/O pins of the device allows the conversion of many channels (in our case one per column) in parallel. The internal processing of the tactile image can also be done in parallel. This means we could be able to respond to very high demanding tasks in terms of dynamic requirements, like slippage detection. This also means we can run complex algorithms at real time, so a smart, programmable and powerful sensor is obtained.

This work presents the development of tactile sensing arrays, inspired by cutaneous sensing in humans, for the fingertips of a humanoid robot. The tactile sensing arrays have been developed in two phases. Microelectrode arrays (MEA), having 32 sensing elements - each epoxy adhered with 25μm thick piezoelectric polymer (PVDF-TrFE) film, were fabricated in the first phase. When connected to the gate of FET devices (external to the chip), each element on MEA acts like an extended gate; thereby facilitating modulation of charge in the induced channel by the charge generated in PVDF-TrFE film - as a result of applied force. Thus, each sensing element converts force into voltage. The tactile sensing arrays developed in second phase work on the same principle but are free from any extended gate. These arrays (having 25 sensing elements) use POSFET (Piezoelectric Oxide Semiconductor Field Effect Transistors) touch sensing elements - in which, piezoelectric polymer film is directly spin coated on the gate area of the FET devices. Thus, a POSFET touch sensing element 'senses and partially processes at same site' - as is done by receptors in human skin. The spatial-temporal performance of these chips is similar to that of skin in the human fingertips.

Epilepsy is the most common chronic disorder of the nervous system. Generally, epileptic seizures appear without foregoing sign or warning. The problem of detecting a possible pre-seizure state in epilepsy from EEG signals has been addressed by many authors over the past decades. Different approaches of time series analysis of brain electrical activity already are providing valuable insights into the underlying complex dynamics. But the main goal the identification of an impending epileptic seizure with a sufficient specificity and reliability, has not been achieved up to now. An algorithm for a reliable, automated prediction of epileptic seizures would enable the realization of implantable seizure warning devices, which could provide valuable information to the patient and time/event specific drug delivery or possibly a direct electrical nerve stimulation. Cellular Nonlinear Networks (CNN) are promising candidates for future seizure warning devices. CNN are characterized by local couplings of comparatively simple dynamical systems. With this property these networks are well suited to be realized as highly parallel, analog computer chips. Today available CNN hardware realizations exhibit a processing speed in the range of TeraOps combined with low power consumption. In this contribution new algorithms based on the spatio-temporal dynamics of CNN are considered in order to analyze intracranial EEG signals and thus taking into account mutual dependencies between neighboring regions of the brain. In an identification procedure Reaction-Diffusion CNN (RD-CNN) are determined for short segments of brain electrical activity, by means of a supervised parameter optimization. RD-CNN are deduced from Reaction-Diffusion Systems, which usually are applied to investigate complex phenomena like nonlinear wave propagation or pattern formation. The Local Activity Theory provides a necessary condition for emergent behavior in RD-CNN. In comparison linear spatio-temporal autoregressive filter models are considered, for a prediction of EEG signal values. Thus Signal features values for successive, short, quasi stationary segments of brain electrical activity can be obtained, with the objective of detecting distinct changes prior to impending epileptic seizures. Furthermore long term recordings gained during presurgical diagnostics in temporal lobe epilepsy are analyzed and the predictive performance of the extracted features is evaluated statistically. Therefore a Receiver Operating Characteristic analysis is considered, assessing the distinguishability between distributions of supposed preictal and interictal periods.

It is described the architecture of the electronics for the control of a wireless endoscopic capsule with locomotive capabilities and advanced sensing and actuating functions. Special emphasis is done to the description of the driver used for locomotion, which is the most innovative element in the capsule.

Surface pattering engineering techniques are essential to fabricate advanced topographies that can be use to modulate cell and tissue response in bio-materials. Particularly, direct laser interference patterning permits fabrication of repetitive arrays and microstructures by irradiation of the sample surface with coherent beams of light. In this study, we explore the possibilities of this technique to produce advanced architectures on several polymeric substrates. The previously calculated interference patterns using the well known interference theory could be directly reproduced on the polymeric surfaces. Moreover, the cross-section of the structured polymers changes depending on the intensity of the laser beams and number of laser pulses, and photomachinability of polymers is highly influenced by laser wavelength. High absorbance of the polymeric materials at specific wavelengths allows the reduction of the laser intensity required to achieve a determined structure depth. In addition, copolymers of methylmetacrylate-styrene were also studied showing that different types of periodic structures could be obtained depending on laser intensity. The obtained results were compared with thermal simulations by finite element methods as well as classical models.

We report on fabrication of periodic arrays of polyethylene glycol diacrylate (PEG-DA), a biocompatible hydrogel, useful in biomedical applications. The structures were produced by means of multi beam laser interference lithography with both nanosecond (266 and 355 nm of wavelengths with pulses lasting 10 ns) and femtosecond pulsed lasers (800 nm of wavelength and 90 fs laser pulses). Configurations involving two, four and five laser beams were utilized obtaining a wide variety of patterns with different feature sizes in the micrometer scale. Through this technique, we demonstrated the ability to fabricate high feature density patterns over large areas without the use of templates or masks. In addition, resolution and geometrical characteristic of the periodic arrays are discussed as function of pulse duration and laser processing parameters. The photopolymerization nature of the process was also investigated.

Wavelength multiplexing is a new approach for achieving optical super-resolution and overcoming the diffraction limits of an imaging lens. In this paper, we extend this approach to microscopy and present a method based upon wavelength coding that is used to develop, construct and experimentally characterize a new type of optical microscope having no objective lenses. In order to extract the collected spatial information we use spectrometer that is part of our microscopic system.

Flexible devices with integrated micro electrodes are widely used for neuronal as well as myogenic stimulation and recording applications. One main intention by using micro electrodes is the ability of placing an appropriate amount of electrodes on the active sites. With an increasing number of single electrodes the selectivity for signal acquirement and analysis is significantly improved. The further advantage of small and elastic structures inside the biological tissue is the perfect fit. This lead to lower traumatisation of the nerve and muscle fibres during and after acute and chronically surgery. Different designed and structured flexible micro electrodes have been developed at the IBMT based on polyimide as substrate material over the last years including cuff, intrafascicular and shaft electrodes. All these systems are generally built up as single sided devices which reduce the possible electrode site half the area. Especially for shaft and intrafascicular applications having double sided electrode arrangement would increase the selectivity enormous. So areas on both sides can be monitored simultaneously. Recent developments of double sided flexible electrode systems lead to promising results especially for varied signal recording. Though these developments revealed some challenges in the field of micromachining including low yield rates. In this work we describe a new technical approach to develop double sided flexible micro electrode systems with a reproducible high yield rate. Prototypes of intrafascicular and intramuscular electrode systems have been developed and investigated by the means of electrochemical characterisation and mechanical behaviour. Additional investigations have been performed with scanning electron microscopy. We also give an outlook to future in vitro and in vivo experiments to investigate the application performance of the developed systems

Biological neural networks are based upon axonal point-to-point connections which inspire connectionist architecture. As we attempt to engineer ever larger analogues of these neural networks we are forced to multiplex neural signals over time shared paths. This can alter timing of neural information, which is critical in real-time oscillatory networks. Because shared paths induce extra delay due to multiplexing signals, traveling on the channel and passing through routing devices, guaranteeing event arrival deadlines across the communication process becomes crucial. This paper addresses issues related to the guarantee of event timings with arbitrary deadline constraints in real-time distributed spiking neural network systems based on token-ring architecture. To achieve this objective, we propose an integrated method in selecting key system parameters. We show that several parameters must be set carefully if event deadlines are to be satisfied. The token holding time (THT) parameter controls the bandwidth allocation for each node in the token-ring network, and must be set properly to avoid deadline misses. The target token rotation time (TTRT) determines both the speed of token circulation and the network utilization available to nodes. TTRT should also be chosen carefully to ensure that the token circulates fast enough while maintaining a high available utilization. As prove of concept, the proposed method is applied to a multi-board spiking neural network system hosting up to 140 analog neurons spread across 7 circuit-boards. Experimental analysis shows that deadline constraints are guaranteed along with bandwidth allocation fairness when applying the proposed method.

A challenging topic of the lab-on-a-chip research is to implement sorting mechanisms on low cost disposable chips. In many applications, surface acoustic waves (SAW) have recently proven to be a versatile and efficient technique for microfluidic actuation. A SAW is excited by applying a high frequency signal to a piezoelectric substrate. When the wave hits the solid/liquid interface it transmits its acoustic energy into the liquid and a local pressure gradient emerges, leading to surface acoustic streaming. Experiments can be performed directly on the piezoelectric substrate or on a separate glass slide positioned on top of the SAW source. We developed a technique for the accumulation of solid and soft objects in SAW generated microvortices in microfluidic channels. For this purpose, the corner of a rectangular microchannel is irradiated by a wide SAW beam. There, the SAW excites sound waves in the fluid producing a typical acoustic streaming flow pattern which typically exhibits two vortices. Particles injected into the flow are accumulated and dynamically trapped in one of these vortices. After the flow is stopped, the collected particles stay in the position of the vortex. In our experiments, we use open microfluidic channels with functionalized hydrophilic-hydrophobic surfaces on glass substrates as well as closed channels build with the elastomer PDMS via soft lithography. We find that the accumulation efficiency for particles is strongly size dependent. Below a critical radius of 500 nm, particles tend to flow through the vortex and are not captured in the corner. Generally, larger particles can be collected at more moderate SAW power levels compared to smaller particles. Therefore, by adjusting the SAW power level, one is able to collect particles above a designated size. This concept is not limited to solid particles but can also be applied to soft objects like cells.

On the basis of a simulation model for lab-on-a-chip systems, the three approaches of sample loop, hydraulic focusing and dielectrophoretic focusing were investigated with a view to increasing sensitivity. The bonding rate can be increased significantly using sample loop and hydraulic focusing. Both approaches involve a considerable extension of the measurement period, though the target bonding rate can be achieved more quickly with sample loop than with hydraulic focusing. By using suitable analytes which can be deflected using dielectrophoresis, the bonding rate can be significantly increased without any extension of the measurement period.

The detection of specific DNA sequences for the analysis of mutations as well as the detection of proteins gains increasing importance in the field of point-of-care diagnostics. Here, a novel low-cost lab-on-a-chip system for label-free detection of DNA hybridization and protein-protein interaction is introduced. The platform consists of a reader with disposable SPR chips produced by injection moulding. Micro optical elements are integrated into the chip to accomplish a simple connection to the optical reader. Automated, software-controlled reagent handling is achieved by a temperaturecontrolled microfluidic system comprising a syringe pump and a switching valve. The sensing area can be separated into maximum 40 parts for parallel analysis. Patterned functionalization is achieved by inverse micro contact printing. Several application examples, ranging from on-chip DNA hybridization up to the detection of antibodies inside diluted human blood serum, will be demonstrated.