Search

The study of ladder networks made by sequences of directly coupled inductor&ndash, capacitor single cells has led us to discover a new property, which may be of certain interest in the sensor field. In the case of n cells, the n-frequencies vector characterizing each node may allow for the identification of that capacitor (sensor), which has experienced a variation of its nominal value. This localization is possible independently from the observable node of the ladder network as proven by the application of the following multivariate data analysis techniques: principal component analysis and partial least square discriminant analysis. This property can be applied on a large scale down to micrometric dimensions in agreement with the technologic ability to shrink the capacitive sensor dimensions.

This paper takes into consideration a number of different ladder networks showing links with the so-called Golden Ratio φ = 1.6180339⋯ , which is embedded into the Fibonacci sequence. Discreteness and periodicity in electronic ladder networks seem to appear as the main features able to include Fibonacci numbers. For instance, in the telecommunication environment, transmission lines can be studied with a given approximation by simplified discrete ladder networks made by a sequence of passive identical cells, such as R-R; R-C; R-L; C-R; L-R; L-C; C-L; directly and electrically coupled each other. In this context, the so called Golden Ratio frequently appears in particular conditions, without a specific reason, at least to our best knowledge, and its presence, so far, still does represent a sort of mystery. © 2014 IEEE.

This paper deals with an analysis of the different behaviors existing between ladder networks (LN) terminated and not Phi terminated by the characteristic impedance Z0. The main purpose of this work is to find, in both cases, links with the golden ratio and, as a consequence, with Fibonacci numbers. Results of some simulations related to the determination of impedances, node voltages and branch currents are given in order to underline approximation effects on the node voltages and branch currents related to both the presence and absence of Z0.This study has been firstly applied to R-R ladder networks but it is here extended to consider other kinds of LNs, having other types of single cells such as C-C ; L-L whose characteristic impedances have also been determined. Double resistive LN have also been investigated for which impedances in any cell, currents in any branch and voltages in any node have been determined in the two cases of presence and absence of the characteristic impedance and with different ratio between longitudinal and transversal impedances. This study has another not less relevant aim: find one of the possible platform for the modelization of biosystems like DNA and RNA, because these structures even if much more complex look like a LN

At present, mammography is the most effective examination for an early diagnosis of breast cancer. Nevertheless, the detection of cancer signs in mammograms is a difficult procedure owing to the great number of non-pathological structures which are also present in the image. Recent statistics show that in current breast cancer screenings 10%-25% of the tumors are missed by the radiologists. For this reason, a lot of research is currently being done to develop systems for Computer Aided Detection (CADe). Probably, some causes of the false-negative screening examinations are that tumoral masses have varying dimension and irregular shape, their borders are often ill-defined and their contrast is very low, thus making difficult the discrimination from parenchymal structures. Therefore, in a CADe system a preliminary segmentation procedure has to be implemented in order to separate the mass from the background tissue. In this way, various characteristics of the segmented mass can be evaluated and used in a classification step to discriminate benign and malignant cases. In this paper, we describe an effective algorithm for massive lesions segmentation based on a region-growing technique and we provide full details the performance evaluation procedure used in this specific context.

Breast cancer is the main cause of cancer deaths among women in the world. About one million new cases appear every year, and about 25% of them lead to the death of the patient. The best solution is the early detection of suspicious tumoral signs through an effective mammographic screening program. Unfortunately, this kind of images is very difficult to interpret by the radiologists because of its very low contrast, so proper image-processing procedures could help them to achieve better diagnoses. This paper improves and assesses an algorithm, already proposed by the authors, that suggests to doctors the suspicious regions that could contain tumoral masses. The procedure succeeds also in the case of very low contrast because it depends only on the orientation of the gradient vectors in the image but not on their amplitude.

Mammographic images suffer from low contrast and signal dependent noise, and a very small size of tumoral signs is not easily detected, especially for an early diagnosis of breast cancer. In this context, many methods proposed in literature fail for lack of generality. In particular, too weak assumptions on the noise model, e.g., stationary normal additive noise, and an inaccurate choice of the wavelet family that is applied, can lead to an information loss, noise emphasizing, unacceptable enhancement results, or in turn an unwanted distortion of the original image aspect. In this paper, we consider an optimal wavelet thresholding, in the context of Discrete Dyadic Wavelet Transforms, by directly relating all the parameters involved in both denoising and contrast enhancement to signal dependent noise variance (estimated by a robust algorithm) and to the size of cancer signs. Moreover, by performing a reconstruction from a zero-approximation in conjunction with a Gaussian smoothing filter, we are able to extract the background and the foreground of the image separately, as to compute suitable contrast improvement indexes. The whole procedure will be tested on high resolution X-ray mammographic images and compared with other techniques. Anyway, the visual assessment of the results by an expert radiologist will be also considered as a subjective evaluation.

Mammography is the most effective method for the early detection of breast diseases. However, the typical diagnostic signs such as microcalcifications and masses are difficult to detect because mammograms are low-contrast and noisy images. In this paper, a novel algorithm for image denoising and enhancement based on dyadic wavelet processing is proposed. The denoising phase is based on a local iterative noise variance estimation. Moreover, in the case of microcalcifications, we propose an adaptive tuning of enhancement degree at different wavelet scales, whereas in the case of mass detection, we developed a new segmentation method combining dyadic wavelet information with mathematical morphology. The innovative approach consists of using the same algorithmic core for processing images to detect both microcalcifications and masses. The proposed algorithm has been tested on a large number of clinical images, comparing the results with those obtained by several other algorithms proposed in the literature through both analytical indexes and the opinions of radiologists. Through preliminary tests, the method seems to meaningfully improve the diagnosis in the early breast cancer detection with respect to other approaches.

In this paper, a new kind of architectures for the full-flash or subranging Analogue to Digital Converters (ADC’s) is presented. To describe these architectures and explain the intrinsic conversion procedure, we refer to a technique, already presented by the author, which improves the accuracy of a measurement system. This technique is based on the arithmetic of the integer numbers in finite fields and include, as a particular case, the well known “caliper rule”. The said technique is recalled and explained by using some simple measurement examples, in particular length, time and mass measurements. Then the technique is applied to the full-flash and subranging ADC architectures, greatly reducing the number of the resistors necessary to generate the voltage reference scale. Indeed, the required resistors for a n -bits conversion word length are reduced from 2 n to about 3 ∗ 2 n / 2 . This reduction and the particular comparison method involved in the theory allow the realization of full-flash architectures having large word lengths.

The UMTS appear to be the communication system of the next future. In this system all the communication services will be included. This means that the Quality of Service (QoS) will have a great importance. So is envisaged that the management of the system will take in account the momentary system capabilities continually determined in real time. Until now, the quality of the links is determined by end-to-end in line procedures based on Bit Error Rate (BER) measurements. It is clear that an end-to-end procedure cannot detect the worse section of the link. To overcome this inconvenient the more recent trend of QoS management is to realise a distributed monitoring of the whole system. Following this last trend this paper proposes a distributed off-line monitoring of the system based on the analysis of the Quality of Analog Signal (QoAS). This proposed analysis implies a high-speed analog to digital conversion of the Intermediate Frequency signal and a very complex processing. In this paper, the state-of-the-art researches performed in this perspective are described.

The UMTS appear to be the communication system of the next future. In this system all the communication services will be included. This means that the Quality of Service (QoS) will have a great importance. So is envisaged that the management of the system will take in account the momentary system capabilities continually determined in real time. Until now, the quality of the links is determined by end-to-end in line procedures based on Bit Error Rate (BER) measurements. It is clear that an end-to-end procedure cannot detect the worse section of the link. To overcome this inconvenient the more recent trend of QoS management is to realise a distributed monitoring of the whole system. Following this last trend this paper proposes a distributed off-line monitoring of the system based on the analysis of the Quality of Analog Signal (QoAS). This proposed analysis implies a high-speed analog to digital conversion of the Intermediate Frequency signal and a very complex processing. In this paper, the state-of-the-art researches performed in this perspective are described.

The computer program PILA (piecewise-linear analysis) is described. It is specially designed for exploring complicated dynamics of moderately sized piecewise-linear circuits. The DC-analysis is performed either by an adapted version of Katzenelson’s algorithm or by an algorithm designed for solving the linear complementarity problem. If the user chooses the latter algorithm, he will get all solutions of the DC-problem, whereas with Katzenelson’s algorithm usually just one solution is found. The time-domain analysis proceeds by the calculation of eigenvalues of the state matrix in order to obtain a time-symbolic solution in each linear region. This method is particularly fast and accurate, as long as the order of the circuit is not too high. The main errors occur in the determination of the transition between two adjacent linear regions. Particular care has been taken to get maximum transition accuracy. Some examples illustrate the various features of PILA.

Microcalcifications clusters are subtle signs of breast cancer and are very difficult to diagnose in the mammographic images because of their small size and low contrast with respect to the normal breast tissue. In this paper, we propose an automatic procedure for the identification of microcalcifications clusters to aid the radiologist in the detection of this kind of breast cancer signs.

Radiologists that analyze screening mammographic images miss the 10–20% of the diagnosis since this kind of images are very difficult to interpret. In this paper, we present the first step of a CADx (Computer Aided Diagnosis) system that, from the original mammogram, extracts suspicious regions on which the radiologists have to focus their attention. The procedure often successes also in case of very low contrast, because it depends only on the orientation of the gradient vectors in the image but not on their amplitude.

The paper presents a first implementation of a novel tool to assist radiologists in analyzing screening mammographic images. The software named Assisted Breast Cancer Diagnosis Environment (ABCDE) is able to acquire DICOM images (also in presence of Grayscale Softcopy Presentation State). It is designed to assist the doctor, as a second reader, during the various phases of the diagnosis. The program uses some public libraries and implements many new algorithms developed in the last years by the research group.

The paper deals with noise power variation that occurs when Discrete Dyadic Wavelet Transform (DDWT) is applied to signals affected by Wide Sense Stationary (WSS) additive white noise owing to the use of a non orthonormal expansion. An exact relationship between the noise variance in the original signal and the noise variance in the wavelet coefficients at a generic level is derived. This relationship is crucial in the application of wavelet thresholding for signal denoising to properly select the threshold in each subband.

Breast cancer is the most common cause of death among women and the most effective method for its diagnosis is mammography. However, this kind of analysis is very difficult to interpret and for this reason radiologists miss 20-30% of tumors. We propose a module for the segmentation of masses that can be implemented in a complete CADx (Computer Aided Diagnosis) system. In particular, we implement a new version of the region growing algorithm specific for this kind of images and for the constraints on computation time of this application.

The problem of uncertainty representation and propagation in the context of statistically correlated variables is commonly addressed by means of Monte Carlo simulation as recommended in IEC-ISO Guide. Moreover, in a recent literature, fuzzy sets have proved to be a valid alternative in the case of independent variables. Unfortunately, the problem of modelling statistically correlated variables, by means of fuzzy sets, is still an open problem. Since it is well known that T-norms are the natural way of combining fuzzy variables into a nonfuzzy function f, in this paper, we investigate how to generalize the class of T-norms, making it dependent from correlation coefficient in order to emulate different statistical correlation degree among variables. In order to validate the model a comparison with central limit theorem will be accomplished in the case of zero correlation while a practical example will be provided in order to compare the proposed method with Montecarlo simulation and with that obtained by uncertainty propagation described in IEC-ISO Guide.

Uncertainty modelling in measurement represents a crucial task since the final result of a measurement process cannot be expressed by a single value, but by a distribution of values over an interval within which the measurements lie with a given confidence level. A classical Type-1 fuzzy set could be the natural choice for uncertainty model, since a Membership Function (MF) intrinsically embeds the concepts of confidence interval and confidence level. Moreover, from computational aspects, working on fuzzy sets is much more easily than a Montecarlo simulation, that is actually the recommended approach. However, both the approaches need reliable and specific assumptions on the probability distribution of the input variables. So, in many practical cases, a Type-2 fuzzy set is needed in order to overcome this problem. In this paper, we will provide a full description of how a Type-2 MF can be built in order to model the uncertainty of a variable and how to evaluate uncertainty propagation through a generic function. A practical example of this representation will be also provided and compared with a Montecarlo simulation.

Conversion from a residue number system (RNS) to binary is a crucial point in the implementation of modular processors. The choice of the moduli is strictly related to conversion performance and influences the processor complexity. In this paper, a conversion method based on a class of couples of coprime moduli defined as (n-2/sup k/, n+2/sup k/) is presented. The method and the related architecture can easily be extended to a larger number of moduli. Since it is based on very small look-up tables, the proposed method can be implemented by fast and low complex architectures.

Mammography is the most effective method for early detection of breast diseases. However, the typical diagnostic signs, such as masses and macrocalcfications, are difficult to be detected because mammograms are low contrast and noisy images. In this paper, we present an algorithm for mammographic images enhancement and denoising based on the wavelet transform. In particular, we develop an adaptive procedure to perform an optimal denoising using a local iterative fuzzy noise variance estimation. Moreover, the degree of enhancement is adaptively tuned at each scale. The proposed algorithm has been tested on clinical images.

Meteorological forecasting is an important issue in research. Typically, the forecasting is performed at “global level,” by gathering data in a large geographical region and by studying their evolution, thus foreseeing the meteorological situation in a certain place. In this paper a “local level” approach, based on time series forecasting using Type-2 Fuzzy Systems, is proposed. In particular temperature forecasting is inspected. The Fuzzy System is trained by means of historical local time series. The algorithm uses a detrend procedure in order to extract the chaotic component to be predicted.

The space vehicles require more and more processing capabilities. This to obtain different advantages: device flexibility, better employment of the frequency spectrum, improving quality and reliability of transmission, preprocessing of the captured data, and so on. These advantages are obtained by implementing onboard complex algorithms. An advantageous realization of these algorithms can be obtained by using residue number system (RNS) arithmetic. The use of the RNS arithmetic requires a complex conversion from the binary representation to the RNS representation. This work describes a new approach to this last problem. A direct analog to RNS converter (ARC) architecture is proposed. This architecture follows from a new further interpretation of the RNS: instead of an elaboration tool this arithmetic is viewed as a measurement tool. The principle used can be considered an RNS extension of the caliper rule. With respect to other realizations already published on S. Mandyam and T. Stouraitis (1990), T. Stouraitis (1992) and D. Radhakrishnan and A. P. Preethy (1998), also by the author and some co-workers on R. Lojacono et al. (2000), the new proposed architecture presents some advantages.

This paper presents a novel methodology for data fitting using a piecewise rational function. That function is able to satisfy any regularity conditions through the direct setting of some free parameters. Moreover, in any grid point of a defined grid we can impose simultaneous interpolation conditions of both the samples and their derivatives. The value of the interpolation function in any point of the input domain depends only on the closest four grid points in its neighborhood. This property bounds the uncertainty propagation in the case of noisy data. A suitable procedure permits to extend the grid beyond the convex hull, so keeping the described properties of the interpolation without affecting the function in the other regions

The passive intermodulation (PIM) is a well known inconvenience in satellite links. It consists in an unwanted generation of harmonics and resulting intermodulation products in weak nonlinear components crossed by RF currents or in irradiation of these unwanted signals by external materials exposed to a strong RF electromagnetic field. Several studies have been carried out to reduce this negative effect and particular measurement techniques have been employed in pre-launch testing of the whole components of the required payload. The extension of the satellite links to mobile communications, which require more transmitted power and increased receiver sensitivity, renders more and more important the reduction of the PIM effects on the downlinks to obtain a satisfactory level of quality of service (QoS). This paper reports the simulation results of effects of the third order PIM products in a multichannel transmitting system. The simulations are carried out by using time-frequency representations of the unwanted combinations of spurious signals generated by third order PIM products. The results of the above simulations are interpreted and a first attempt of design guidelines is given.

In this paper we present a Residue Number System (RNS) implementation of digital filters to be used for space applications. The RNS is particularly attractive because of the reduced power dissipation with respect to filters realized in the traditional binary representation. Furthermore, the reduced circuit complexity allows the implementation of programmable structures, in terms of filter coefficients and dynamic range, a feature appealing for remote operations on systems onboard satellites. TABLE OF CONTENTS

The paper faces the problem of designing a compact processor devoted to the evaluation of a time-frequency transformation for space applications. The guidelines to save circuitry and power consumption are searched. Starting from the simple Gabor transformation, the effects of the Weyl-Heisenberg indetermination principle are examined in connection with the effects induced by the use of a short wordlength. A simple criterion to choose the optimal wavelet is given and applied to improve the transformations performed using a reduced wordlength.

This paper deals with the design of a digital demultiplexer based on a multirate filter bank. In particular a digital architecture for replacing an analog filter bank composed of six surface acoustic wave (SAW) filters have been developed. This analog structure is used, at present, by Alenia Space Division in the Hot-Bird satellites. An extensive comparison between alternative algorithms for the demultiplexer implementation has been carried out. Moreover different implementations have been compared in terms of the computational complexity and accuracy with respect to the quantization effects. Starting from the simulation results, the obtained architecture has been mapped on six Altera Flex 10 K-100 devices. The final test bed, that includes a complete interface to the Alenia demodulator has been implemented on a four-layer PCB.

The characterization of the quantization effects due to the conversion of infinite precision quantities to fixed-point arithmetic constitutes a relevant research area. It is of fundamental importance for the development of automatic tools for the transformation of DSP algorithms from floating point to fixed point arithmetic. In this paper a novel approach for the direct computation of the probability density function (PDF) at the output of generic memory-less nonlinear processing blocks is derived. Our analysis has been applied to the modeling of the THA-ADC chain. By using this approach, we are able to determine the exact behavior of the Signal to Noise Quantization Ratio (SQNR) for cases where the well-known Pseudo Quantization Noise (PQN) model does not give accurate results.

The application of the residue number system (RNS) to signal processors has been widespread. In the case of infinite impulse response (IIR) digital filters, the application of the method implies the continuous scaling of the results, as in the case of autoscale multipliers. A different method of applying the RNS technique to fixed-point arithmetics was previously proposed by the authors. For this purpose, two finite arithmetics defined with respect to two different moduli were used. A case of particular interest was considered, i.e. that of 8-bit arithmetic, in which a limited amount of memory is needed to store the isomorphism tables. An extension of the above algorithm to the case of higher dynamic ranges is considered. The approach uses a radix-p representation and requires the same isomorphism tables already defined in the case of 8-b arithmetics. The algorithm thus preserves the memory requirements. If p=128 and the wordlength is 7L+1, the 8-b processor can be used as an elementary building block of the whole multiplying structure. A 22-b multiplier is considered in detail and compared to the traditional ones with respect to the time and area requirements for a VLSI realization. >

The authors propose a residue number system (RNS) architecture that realizes fast division. This architecture is fully developed in the particular case of eight-bit wordlength; in this case significant simplifications are reached. The whole divider appears as a combinatorial structure. A layout of the proposed architecture has been realized in a 1.5- mu m CMOS technology, SPICE simulation of the circuit extracted from this layout has shown the overall propagation time in performing the division to be about 100 ns. >

The computer program PILA, which is designed for the fast and accurate analysis of moderately-sized piecewise-linear circuits, is described. Since PILA does not use numerical integration, it is particularly useful for exploring the qualitative behavior of circuits with complicated time evolutions, e.g. in the presence of chaotic solutions. The computational accuracy and the capability to deal with linearly dependent capacitor charges and inductor fluxes are highlighted with two examples. >

Digital signal processing using the residue number system (RNS) is an area of renewed and growing research interest. A basic difficulty is the division required by most of the adaptive algorithms. The authors propose a method to implement a unified RNS architecture for both product and general division. The architecture of a whole RNS Burg cell and the layout of the main blocks in a 1.5 mu m p-well CMOS technology are presented. The RNS realization is compared with a classical NMOS implementation previously developed by the authors. >

In this paper a very simple VLSI architecture of digitally programmable OTA-C filter is presented. The proposed realization uses a Digitally Programmable Transconductance Amplifier (DPTA) that permits to vary its transconductance parameter in digital way. Thus the tuning of many features of biquad section is obtained simply through digital control lines.

The possibility of designing RNS (residue number system) recursive digital filters with reduced internal wordlength is considered. For this purpose, a class of RNS filters is introduced which are not affected by coefficient and signal overflows in internal points. The proposed filter architecture consist of two independent and parallel modular processors. A complete example of RNS fifth-order low-pass filters with 8-b wordlength is developed and compared with conventional fixed-point implementations. The proposed algorithm exhibits arithmetical features similar to those of conventional ones with larger wordlengths. This result is very promising in view of the development of ASIC architectures. >

A technique called the double module method (DMM) is applied to recursive digital filters. The method confines overflow and quantization errors to some internal points of the filter, so that behavior even better than that of conventional fixed-point realization can be obtained for a given word length. Some considerations regarding the VLSI realization of the filters are included. >

The scaled Chinese Remainder Theorem (CRT) is a very useful tool for the simplification of RNS to binary converters. The main drawback of this methodology is related to the use of large look-up tables that store the correspondence among the modular numbers and the corresponding scaled terms of the CRT. This fact limits the maximum speed allowed by this approach. In this paper a new method for the computation of the scaled factors is presented. It allows the computation of the scaled CRT output by using very small look-up tables implemented by conventional logic and simple and fast structures, that work in parallel. The only assumption made in order to develop the new algorithm is that the moduli must be odd.

A fast architecture for the implementation of FIR filters based on the residue number system is proposed. High-speed is obtained by introducing residue arithmetic that permits the computation of the filter output by using N FIR subfilters of reduced dynamic range operating in parallel. Moreover, a new approach for the computation of the modulo operation on each term of the linear combination has been developed. The filter implementation is based on small look-up tables and fast multioperand carry-save adders.

RNS to binary conversion is a very critical task for an advantageous implementation of RNS based Digital Signal Processors (DSPs). This paper presents a new technique for the module extraction useful for the implementation of efficient Chinese Remainder Theorem (CRT) based converters. The method, based on the multiplication of the CRT terms by a power of two, greatly simplifies the extraction of the final modulo only requiring a look-up table access and two sums. The resulting structures appear very efficient in terms of speed and area occupancy.

In this paper we describe the most important steps of the VLSI analog implementation of a learning actractor neural network. An overview of the theoretical model is presented while the most important circuits are described in greater details.

This paper presents a novel methodology for the analysis of the downlink of a DS-CDMA mobile satellite system which allows one to evaluate the impact of the power limitations expected in a satellite-based access in terms of capacity. The methodology has been applied to compare an optimum power control, providing BER-driven thresholds, with a pure signal-to-interference ratio (SIR)-based strategy. Relationships among service BER, channel states statistic and expected dynamic range of the outer-loop process are derived. Concluding remarks on the power management policy, and its impact on the overall capacity, are given.

Some possible applications of the time-frequency representation (TFR) of signals in the management of the quality of service of communications systems are described. After a brief review of some of the numerous available transformations, attention is particularly focused on linear transformations. The linear relationship between the signals in the time domain and the representations in the mixed time-frequency domain facilitates actions to improve QoS management. An example applied to the detection of multipath effects is given.

In this chapter, some methodologies for developing fuzzy hardware architectures are presented. The architectures can be grouped into two sets, namely general-purpose structures, and task oriented structures. We discuss the effectiveness of the task-oriented architectures for the implementation of fuzzy processors. We present an architecture exploiting the properties of the fuzzy processing in order to simplify and speed-up its basic blocks. In particular, the fuzzification unit is based on a suitable class of trapezoidal membership functions with power of two slopes for the oblique sides. The algorithm used for the defuzzification unit is based on the Center of Gravity (COG) method. It is implemented by a full look-up table approach in order to obtain the maximum speed. With respect to other look-up table approaches the architecture we propose usesess memory and is fully parallel. In addition, a divider that computes an approximate result has been added and further reduces the overall complexity of the circuit. The structure has been designed and simulated by using the ES2 technology. We obtained a chip of 7 mm2 for the core logic, and an external 10 Kbytes memory containing the output membership functions. The simulations have shown a speed over 100 KFLIPS for 2 inputs and 20 rules. The effectiveness of this architecture has been tested on a specific high-speed application, i.e. the radar target tracking. In this case, the simulation results give good performance in terms of target tracking capability and speed higher than 400 KFLIPS.

The scaled Chinese remainder theorem (CRT) is a very useful tool in residue arithmetic. Its properties can be exploited for the simplification and speeding-up of the conversion process. The main drawback presented by this methodology, when it is used for the output conversion, is the need of long wordlength look-up tables (LUTs) storing the correspondence among the modular numbers and the corresponding scaled terms of the CRT. This fact limits the maximum speed obtainable by this approach. In this brief, a new method for the computation of the scaled terms is presented. It has been implemented by using very small wordlength LUTs and simple arithmetic operators. The only proviso is that the moduli must be odd. The obtained architecture is very fast and due to the local interconnections is suitable for an efficient VLSI implementation.

Cellular Neural Networks (CNN's) have been introduced by L. O. Chua and L. Yang as a class of large-scale non-linear analog circuits which process signals in real time. It is well known that these networks are useful in a number of impressing applications, as in the field of image processing. A very favourable property of CNN is to have only local interconnection links among cells, which makes them very suited for VLSI implementation. The main purpose of this paper is to remove the limitation of a fixed cloning template implemented in some previous chips, realizing VLSI hardware of reconfigurable CNN's. In this way, the cloning template is not inherent to the hardware but it can be updated by an external control line through a D/A converter. A complete array of 8x8 analog CNN cells and the hardware of the whole control circuitry is realized in a single-chip.

A residue number system (RNS) method to implement fast recursive digital filters is considered. The speed improvement is obtained as a better performance of the elementary arithmetic blocks, given by the RNS. In the past the RNS double moduli method has been applied as a scaling procedure. In the present work a technique which uses this method partially is presented to improve the performance of recursive filters. This technique is based on the Chinese remainder theorem (CRT). Using CRT a partially parallelized structure can be realized which reduces the computational time and improves the speed performance. The effectiveness of this method has been tested numerically and by a 1.5- mu m CMOS implementation of a second order infinite impulse response (IIR) filter cell. >

The residue number system (RNS) digital circuit is defined as a circuit constituted by delays, modular adders, and multipliers. The synthesis of RNS digital circuits is considered starting from a given impulse response. The key requirement of the proposed synthesis procedures is the length of the sequence, which depends on the modulus, the type of the poles, and their multiplicity. The definitions of the different types of poles require the use of suitable extensions of the basic field corresponding to the modulus. >

The possibility of realizing structural passive digital filters in terms of residue number systems (RNS) is considered. For this purpose, an algorithm is proposed for realizing the RNS rotator, which represents the basic element of these filters. The algorithm is discussed and illustrated in detail in the case of an 8-b dynamic range. The possibility of extending the dynamic range by the Chinese remainder theorem (CRT) is also considered and a simplified reconstruction formula is proposed for reducing the computational cost of the CRT. >

A novel approach to the definition of the roots of unity in the RNS (residue number system) is presented. The definition is based on a suitable polar representation of the complex residues. The resulting RNS Fourier transform provides an algorithm for performing circular convolutions characterized by flexibility in terms of length, computational cost, and storage. >

Meteorological forecasting is an important issue in re- search. Typically, the forecasting is performed at "global level", by gathering data in a large geographical region and by studying their evolution, thus foreseeing the meteorological situation in a certain place. In this paper a "local level" approach, based on time series forecasting using Type-2 Fuzzy Systems, is proposed. In particular temperature and humidity forecasting is inspected. The Fuzzy System is trained by means of historical local time series. The procedure uses a frequency signal decomposition in order to extract chaotic and de- terministic components, so as they be independently predicted. This technique allows us to achieve more accurate forecasting results even many hours in advance.