Search

Cellular neural networks show high performance capabilities in real time image processing applications. For this reason, their use in biomedical image analysis can be a useful aid to the doctor in clinical diagnosis. In this research area the improvements in systems for clinical specular microscopy in vivo made a strong contribution to the study and the comprehension of the physiopathology of corneal endothelium. The more recent systems allow acquisition of the images and morphometric analysis. Nevertheless, the results (i.e. the automated reconstruction of the endothelium cell borders) are often inaccurate. Moreover, they do not allow the correct recognition of the cell shapes. On the other hand, even if the semiautomatic systems allow an effective evaluation of the cell shape, they are highly time consuming and provide results that could be affected by the criterion used by the operator in the cell corner detection. In this paper a software tool for the full automated morphometric analysis of corneal endothelium images is presented. The tool makes use of an analogue cellular neural network algorithm that allows both cell shape recognition and endothelial cell area measurement.

Cellular Neural Networks (CNN's) represent a remarkable improvement in hardware implementation of Artificial Neural Networks. In fact, their regular structure and their local connectivity feature make this class of neural networks really appealing for VLSI implementations. The CNN are widely used in several application fields, such as image processing and pattern recognition. In this research area, the authors presented a fully digitally programmable CNN chip with 6/spl times/6 cells (6/spl times/6DPCNN chip). In this paper, a system with twenty of these chips will be presented. This system is made up of twenty boards with one 6/spl times/6DPCNN chip each. Its main features are: fully programmability of the templates; digital input/output for logic operation; analog outputs for dynamic analysis; implementation of space-variant as well as space-invariant CNN.

In environment sensing, one of the most relevant knots is certainly represented by the implementation of an effective system for the real time extraction of the three-dimensional information. Several effective approaches on this topic have been proposed in literature. Among the others, the Stereo Vision approach seems to be really attractive. Moreover, such an algorithm can be effectively mapped on a Cellular Neural Network. Consequently, it can be analogue implemented overcoming to the difficulties leading from the relative high computational cost and the required real time processing. In this paper, a new CNN chip well suited for a Stereo Visual System will be presented.

The implementation of an artificial vision algorithm in real time is really attractive in such an application as the field of environment sensing. The SVCNN (stereo vision cellular neural network) chip is an analogue circuit able to compute in real time the Disparity Map from a couple of images by using a stereo visual system algorithm. A "test-bed" board for the 16/spl times/64 SVCNN chip is presented in this paper. This board is composed of an analogue processing core implemented by two 16/spl times/64 SVCNN chips together with a digital high performance pre-processing unit and a video grabbing section.

A new learning algorithm is proposed for space-varying cellular neural networks, used to implement associative memories. The algorithm exhibits some peculiar features which make it very attractive: the finite precision of connection weights is automatically taken into account as a design constraint; no multiplication is needed for weight computation; learning can be implemented in fixed point digital hardware or simulated on a digital computer without numerical errors.

In this paper the hardware implementation is discussed of a cellular nonlinear network (CNN) for analog simulation of reaction-diffusion partial differential equations. The elementary cell consists of the classic nonlinear circuit typical in these CNNs, except for the characteristic of the nonlinear resistor; that is chosen to be non-monotone. The control laws describing the contribution of the neighbor cells depend both on the currents and the voltages of the capacitors.

The real-time estimation of the distance of objects from an observer is a critical issue in several application fields. A new cellular neural network circuit that uses a stereo vision algorithm to compute the disparity map is presented.

One of the most essential requirements in robotic autonomous navigation is the extraction of three-dimensional information about the environment in order to avoid collisions with moving or fixed obstacles. Among the others, one of the most promising approaches for this task is represented by the techniques of artificial vision. Several implementations of different approaches have been proposed in many papers in literature. In particular, the authors presented an implementation of the stereo vision algorithm using cellular neural networks. In this paper, the design of an electronic board with dedicated CNN analogue chips able to implement the algorithm is presented.

The correct classification of single musical sources is a relevant aspect for the source separation task and the automatic transcription of polyphonic music. In this paper, we present a classification experiment on six different musical instruments: violin, clarinet, flute, oboe, saxophone and piano. It is characterized by two steps. In the first step, a suitable signal preprocessing based on FFT and QFT (Q-constant Frequency Transform) is adopted for feature extraction and data set preparation. In the second step, a nonexclusive classification method is proposed to handle the inevitable overlapping among classes. It is obtained by a co-operative clustering technique. The success of this kind of classification method is conditioned by the adopted clustering procedure. We propose a hierarchical scale-based approach for this task, carrying out good results.

Obstacle avoidance is the main issue in autonomous robotics. It requires a three-dimensional effective environment sensing in real time. Among the others, the stereo vision approach to environmental information extraction seems to be very appealing, even if it leads an extremely high computational cost. However, a high performance implementation of this algorithm on a cellular neural network is able to overcome these difficulties. In the paper, the design of a CNN chip well suited for this algorithm is presented. This chip, performing a real time processing of the stereo vision data, will improve the cruising speed of a robotic platform.

Several CNN hardware implementations have been presented in the last year. Most of them consist of one-chip circuits. Among the others, the authors presented the DPCNN chip family: a current-mode interconnection-oriented CNN analogue chip family. This approach presents many advantages and some drawbacks. In fact, simply connecting together more of these chips, it allows one to implement any size CNN arrays with the possibility to modify the network topology. On the other hand, it requires a high number of pads for the interconnections. Moreover, some parasitic capacitors due to the pads and the PCB wiring will appear. In this paper a high performance technique to improve the interconnection strategy able to overcome these drawbacks without any lack of functionality is presented.

The high parallel analogue processing rate makes the cellular neural networks paradigm really useful in such a problems where real-time replies to external stimuli are required. The development of an effective system for autonomous robot navigation can find a valid support from this research. Moreover, the growth of new CNN algorithms can afford the necessary feedback to the hardware developers to improve their realisations. In this paper some measurements of a stereo-vision algorithm on a CNN hardware implementation (the 720DPCNN system) are given.

The design and realisation of reliable hardware CNN systems with a high number of cells is a key point in research in this field. On this topic, several different solutions have been proposed in VLSI implementations. In previously published papers, the authors presented current-mode interconnection-oriented integrated circuits to realise wide CNN networks. The multi-chip architecture shows the drawback to be a pad-limited structure because of the growing number of the pads required by the interconnections. In this paper a technique to improve the interconnection architecture without any lack of functionality will be presented. This approach will drastically cut the interconnection requirements by 75%. Some simulation results are presented together with experimental results.

Cellular neural networks are a remarkable artificial neural network class well suited for real time image processing tasks. In fact, the parallel analogue computing feature makes them really effective in such problems which require a real time response. Moreover, the limited amount of interconnections relative to cell's neighbourhood only, lend themselves to easy VLSI implementation. In previous papers, the authors presented some CNN hardware. Therefore, in this paper, an algorithm for character recognition developed on the 9/spl times/9 DPCNN board is presented.

Among artificial neural network architectures, the Cellular Neural Networks represent one of the most attractive from a VLSI point of view. The 3/spl times/3 DPCNN chip, formerly presented by the authors, is a very effective implementation of CNN's. This paper presents a multichip 9/spl times/9 CNN board which is made up of nine 3/spl times/3 DPCNN chips. The board is fully programmable by a Personal Computer which enables the selection of the parameters of the network and the steady-state voltages acquisition. The system PC/9/spl times/9CNN board represents a very powerful tool in the investigation of new CNN algorithms as well as their dynamic behaviours.

Cellular neural networks represent one of the most interesting neural network class for VLSI implementations. Their main feature is the local interconnections among cells that is really appealing for hardware realizations. This paper presents a multichip 9/spl times/9 CNN board which is made up of nine 3/spl times/3 DPCNN chips. This 9/spl times/9 CNN board is fully controllable by a Personal Computer. A proper software program is presented which enables the selection of all the entries of the templates, the input and initial voltage, and the steady-state voltage acquisitions.

The implementation of a versatile VLSI chip represents an important step to develop cellular neural networks (CNN). In this paper a VLSI realization of the multi-chip oriented, 6/spl times/6 digitally programmable cellular neural network (6/spl times/6 DPCNN) chip, is presented. This chip covers most of the available one-neighbourhood templates for image processing applications. Moreover, it can be easily interconnected to others to form very large CNN arrays.

Cellular neural networks represent a very powerful tool in image processing applications. Moreover, several interesting CNN's applications as logic operators are possible too. Although this is true, their successful development is strongly dependent on a proper hardware implementation. This paper presents some measurement results on a 3/spl times/6 CNN made up of two 3/spl times/3DPCNN chips implementing logic operations.

In this paper a high performance VLSI implementation of a 3/spl times/3 Digitally Programmable Cellular Neural Network with discrete templates is presented. This chip, manufactured and successfully tested, gives an efficient solution to the hardware implementation of the Cellular Neural Networks. Moreover this chip can be connected to others to carry out very large CNN arrays. This implementation covers the 66% of the available one-neighborhood fixed templates for image processing applications. >

A method for grey-scale image rotation using cellular neural networks (CNNs) is proposed. The rationale of the method is the conversion of image rotation into a small rotation followed by a sequence of translations of pixel blocks. The same programmable 3×3 space-varying template can be used in both cases.

A new IIR adaptive filter for real-time, room acoustic response simulation is proposed, the structure of which derives from Jot's model of an artificial reverberator. The simultaneous perturbation stochastic approximation (SPSA) algorithm is used to set parameter values. Results show good similarity between the desired and artificial response.

A major problem in the VLSI implementations of cellular neural networks (CNNs) is that of achieving ease of programmability of the templates. The authors present a very efficient VLSI implementation of a digitally programmable transconductance amplifier for cellular neural network applications with discrete templates.

In this work a high frequency model of a photovoltaic (PV) plant is identified and implemented aiming to investigate the common mode (CM) currents circulating through the ground connections of the plant. From the identification of the impedance obtained by frequency response of photovoltaic module and by a suitable of the power conversion unit, an equivalent high frequency circuit representation has been developed. The lumped parameters circuit model is implemented in PSpice environment to obtain the CM leakage currents. Harmonics at frequencies multiple of the switching frequency and a strong resonance at about 12 MHz are detected.

The switching power converters are used in a broad variety of applications, from the consumer electronics to the DC distribution systems, from the vehicle applications (road vehicles, marine vehicles, aircraft) to the industrial automation. In each of these application fields, the conversion systems which present more compact size and reduced weight, at the same power, are strongly required in relation to stringent design constraints. In this context, the optimization of the power density of the converter becomes an essential requirement. The increase of the switching frequency of the static devices allows an improvement of the power density, thanks to the possibility of reducing the sizes of the energy storage passive components (inductors and capacitors). On the other hand, the increase of the switching frequency determines, with high probability, the generation of more relevant conducted electromagnetic interferences (EMI) in the frequency range 150 kHz – 30 MHz. In particular, the high switching frequency is responsible for several serious problems affecting both the reliability and the electromagnetic compatibility of the systems of which the converter is part. For this reason, noise mitigationg is, more than ever, one of the major issues in power electronic system design, particularly when dealing with stringent standard regard the maximum emission limits, which however are mandatory for the marketing of these systems. EMI filters are the most efficient among the different possible solutions to mitigate the conducted electromagnetic interferences. On the other hand, EMI filters are part of the power electronic converters and they have significant impact on the overall converter volume and weight. In order to take on this issue, besides satisfying EMI limits, a further optimization in terms of filter size and weight during the design stage is advantageous to maximize the overall converter’s power density. The identification of the configuration leading to the best power density, in terms of minimum volume/weight, is a nontrivial task. The conventional design of EMI filters disregards the power density issue. The trial and error approach requires a significant effort in terms of time spent and it does not guarantee the optimal choice of filter configuration in order to obtain the maximum power density. For this reason, an automatic optimized design procedure of discrete EMI filters has been developed. Once the power electronic converter characteristics are known and based on databases, suitably set up, of commercially available devices for the realization of EMI filters, the optimized procedure enables EMI engineers or scientists to obtain the best EMI filter configuration in term of power density. On the basis of the developed automatic design procedure, an interactive software, ODEF (Optimized Design of EMI Filters), has been developed to make the new design procedure more accessible to EMI designer. Moreover, the developed application is provided of a graphical interface which allows to analyze and compare simultaneously different EMI filter designs. The optimization algorithm can be used as a EMI filter design tool but also as a tool for the analysis of the EMI filters performance.

In this work, a previous software used to simulate partial discharges (PDs) under Alternating Current (AC) stress has been modified in order to evaluate the PDs behavior under a voltage stress close to the Direct Current (DC) waveform. By using a full-wave and a half-wave rectifier, the specimen with an air void defects has been subjected to a gradual constant stress. Finally, a capacitive filter has been inserted in order to produce a steadier voltage supply. Simulation results show that under an almost DC waveform, the PDs activity become less compared to AC stress.

The paper deals with low-frequency magnetic field measurements carried out by using a broadband and isotropic instrument. These measurements are characterized by very high uncertainty values, which imply a high risk of wrong decisions when there is the need to establish if a site complies or does not comply with specified emission limits. To reduce this risk, we decided to perform the so called “uncertainty management” that is the discipline of optimizing the cost of a measurement versus the uncertainty target. The task is achieved by using the PUMA method that is an iterative technique originally conceived for geometrical and mechanical measurements. The approach is completely based on the “Guide to the Expression of Uncertainty in Measurement” rules, but provides a more engineering methodology. By using this approach, it is possible to avoid the usage of too expensive instrumentations or, on the contrary, too expensive resources for the uncertainty estimation.

The expression of emotion is an inherent aspect in singing, especially in operatic voice. Yet, adverse acoustic conditions, as, e. g., a performance in open-air, or a noisy analog recording, may affect its perception. State-of-the art methods for emotional speech evaluation have been applied to operatic voice, such as perception experiments, acoustic analyses, and machine learning techniques. Still, the extent to which adverse acoustic conditions may impair listeners' and machines' identification of emotion in vocal cues has only been investigated in the realm of speech. For our study, 132 listeners evaluated 390 nonsense operatic sung instances of five basic emotions, affected by three noises (brown, pink, and white), each at four Signal-to-Noise Ratios (-1 dB, -0.5 dB, +1 dB, and +3 dB); the performance of state-of-the-art automatic recognition methods was evaluated as well. Our findings show that the three noises affect similarly female and male singers and that listeners' gender did not play a role. Human perception and automatic classification display similar confusion and recognition patterns: sadness is identified best, fear worst; low aroused emotions display higher confusion.

In this paper, an experimental study, carried out on an operating MV cable, is presented. The aim is to identify, locate and characterize defects in the cable's EPR (ethylene-propylene copolymer rubber) insulation. Direct partial discharge measurements have been compared to those obtained by high frequency electromagnetic field measures. The latter have been carried out using a sensor which does not require any galvanic connection with the system under test and that performs an intrinsic localization of the defect. Moreover, an analysis of an MV cable junction has been carried out using the same tools.

The VLSI implementation of Cellular Neural Networks is a relevant task which is very important for the future development of neural networks. In this paper a a modular VLSI implementation of a 3/spl times/3 Digitally Programmable Cellular Neural Networks is presented. This chip is the first successfully tested fully programmable Cellular Neural Network hardware implementation. It covers most of the available one-neighborhood templates for image processing applications. Moreover, it has been designed to be easily interconnected to others to give very large CNN arrays.

In this paper the performances of rainfall energy harvesting by means of piezoelectric transducers is presented. Diverse studies agree on the level of suitable generated voltage on the electrodes of a piezoelectric transducer subjected to rainfall, but a complete characterization on the supplied power is still missing. This work, in order to limit optimistic forecasts, takes into account the behavior of the transducers subjected to real and also artificial rainfall, condition that has shown promising behavior in laboratory. In order to increase the energy harvesting and also define its limits different loads have been taken into account. Only commercial transducers have been considered: a lead zirconate titanate and polyvinylidene difluoride transducer.

The partial shading phenomenon is a well known problem of photovoltaic plants. Partial shading leads to undesirable effects such the electrical mismatch, the generation of hot spots, and generally the decrease of production of electric energy. To mitigate the last effect, a dynamic reconfiguration of the electrical connections between modules was taken into account. In this paper, starting from an already developed system for a small-scale photovoltaic plant reconfiguration, the study of the economical benefits of the employment of a reconfigurator are traced. Five different incentive policies of diverse Countries have been considered to evaluate the increase of Net Present Value of system with and without a reconfigurator.