This paper gives an overview of recent progress in radar polarimetry and radar polarimetric interferometry. Both techniques are of special importance for the inversion of physical scatterer parameters from radar remote sensing data. A unified treatment of polarisation effects in radar polarimetry and polarimetric interferometry based on eigenvalue processing is addressed providing a link between signal processing techniques and coherent electromagnetic models for random media scattering. In this context, the main applications of polarimetry in radar remote sensing such as single and multi-frequency polarimetric classification, the estimation of surface roughness and moisture content and vegetation structure estimation are reviewed.

In this article, a hierarchical classifier is proposed for classification of ground-cover types of a satellite image of Kangaroo Island, South Australia. The image contains seven ground-cover types, which are categorized into three groups using principal component analysis. The first group contains clouds only, the second consists of sea and cloud shadow over land, and the third contains land and three types of forest. The sea and shadow over land classes are classified with 99% accuracy using a network of threshold logic units. The land and forest classes are classified by multilayer perceptrons (MLPs) using texture features and intensity values. The average performance achieved by six trained MLPs is 91%. In order to improve the classification accuracy even further, the outputs of the six MLPs were combined using several committee machines. All committee machines achieved significant improvement in performance over the multilayer perceptron classifiers, with the best machine achieving over 92% correct classification.

Tree canopies contain various scattering elements such as leaves, branches and trunks, which contribute to complex backscattering, depending on frequency and polarization. In this paper, we propose to use the polarimetric correlation coefficient for classifying trees, forests, and vegetations. The polarimetric correlation coefficient can be derived by the elements of Sinclair scattering matrix. Since the scattering matrix can be defined in any polarization basis, we examined the coefficient in the linear HV, circular LR, and optimum polarization bases. First, the change of correlation coefficient inside trees along the range direction is examined using small trees in a laboratory. The wider the range, the better the index. The coefficient defined in the LR polarization basis showed the largest change within tree canopy, which also contribute to retrieve scattering mechanism. Second, this index for discrimination is applied to polarimetric SAR data sets (San Francisco and Briatia area) acquired by AIRSAR and SIR-C/X-SAR. It is shown that polarimetric correlation coefficient in the LR basis best serves to distinguish tree types.

We present a simple cell scheduling algorithm for an input buffered switch. The suggested algorithm is based on iSLIP and consists of request, grant and accept steps. The pointer update scheme of iSLIP is simplified in the suggested algorithm. By virtue of the new update scheme, the performance of the suggested algorithm is better than that of iSLIP with one iteration. Using computer simulations under a uniform traffic, we show the suggested algorithm is more appropriate than iSLIP for scheduling of an input buffered switch with multiple service classes.

A class E low dv/dt PWM synchronous rectifier regulating the output voltage at a fixed frequency is presented, analyzed and verified experimentally. This rectifier is derived from the class E low dv/dt rectifier by replacing the controlled switch (MOSFET with its anti-parallel diode) with the rectifier diode in class E low dv/dt rectifier, and by using the synchronized PWM signal to control the output voltage at desired value. The ZVS condition of the controlled switch can be maintained from full-loaded to open-loaded. The experimental results measured at switching frequency 1 MHz are in good agreement with the theoretical prediction.

The design of artificial neural networks (ANNs) through simulated evolution has been investigated for many years. The use of genetic algorithms (GAs) for such evolution suffers a prominent problem known as the permutation problem or the competing convention problem. This paper proposes a new crossover operator, which we call the selected node crossover (SNX), to overcome the permutation problem of GAs for evolving ANNs. A GA-based evolutionary system (GANet) using the SNX for evolving three layered feedforward ANNs architecture with weight learning is described. GANet uses one crossover and one mutation operators sequentially. If the first operator is successful then the second operator is not applied. GANet is less dependent on user-defined control parameters than the conventional evolutionary methods. GANet is applied to a variety of benchmarks including large (26 class) to small (2 class) classification problems. The results show that GANet can produce compact ANN architectures with small classification errors.

In this paper, we deal with the problem of compatibility class encoding, and propose a novel algorithm for finding a good functional decomposition with application to LUT-based FPGA synthesis. Based on exploration of the design space, we concentrate on extracting a set of components, which can be merged into the minimum number of multiple-output CLBs or LUTs, such that the decomposition constructed from these components is also minimal. In particular, to explore more degrees of freedom, we introduce pliable encoding to take over the conventional rigid encoding when it fails to find a satisfactory decomposition by rigid encoding. Experimental results on a large set of MCNC91 logic synthesis benchmarks show that our method is quite promising.

Improvements of Internet technology during the last decade have shifted the technical focus from reachability to the quality of communication. There are many technical frameworks, such as Integrated Service and Differentiated Services, which have been standardized to assure the quality of communication. QoS routing is also one of such frameworks. It changes or fixes a route that IP datagrams take, and is also indispensable to put a variety of services into practice. Nevertheless, experiment reports of QoS routing on operational network are quite few, especially with dynamic SLA. Therefore, we still do not know much about the important factors for QoS-enabled network to be realized, such as users' behavior, suitable services to offer, and configuration parameters. In this paper, we carried out field-trial with pseudo QoS routing and dynamic SLA in an actual network built at the WIDE retreat in autumn 2000. In this field-trial, we provided two different types of links to attendees. Attendees chose one of the links, through which their flows go, with our dynamic SLA. We describe the details and the results of this experiment. Our results could help to understand the customers' behavior for differentiated services, and therefore be useful for designing and deploying various QoS technologies.

This paper describes a method of age-group classification of young males based on their facial images. The facial shapes of males and females are mostly formed by age 20 and 15, respectively. Our study only considered young males as they have a longer period during which facial shape is a determining factor in age estimation. Age classification was carried out using artificial neural networks. We employed 440 facial images in our experiment, composed of 4 different photographic images taken at ages 12, 15, 18 and 22 of 110 young males. Two methods of age classification were used, each employing different features extracted from the facial images, namely, "mosaic features" and "KL features. " As a result, we obtained about an 80% successful classification rate using mosaic features, and a slightly lower rate using KL features. We also analyzed the connection weights between the hidden and input layers of the trained networks, and examined facial features characteristic to each age group.

This paper presents a new approach to modeling speech spectra and pitch for text-independent speaker identification using Gaussian mixture models based on multi-space probability distribution (MSD-GMM). MSD-GMM allows us to model continuous pitch values of voiced frames and discrete symbols for unvoiced frames in a unified framework. Spectral and pitch features are jointly modeled by a two-stream MSD-GMM. We derive maximum likelihood (ML) estimation formulae and minimum classification error (MCE) training procedure for MSD-GMM parameters. The MSD-GMM speaker models are evaluated for text-independent speaker identification tasks. The experimental results show that the MSD-GMM can efficiently model spectral and pitch features of each speaker and outperforms conventional speaker models. The results also demonstrate the utility of the MCE training of the MSD-GMM parameters and the robustness for the inter-session variability.

A new scheduling algorithm called the Adaptive Weighted Round Robin with Delay Tolerance (AWRR/DT) is presented. This scheme can adapt to the traffic fluctuation of networks with a small processing burden. The proposed scheme incorporates a cell discarding method to reduce the QoS degradation in high-loaded (or congested) period. Simulation results show that the proposed scheme can reduce the average delay of the non-real-time (NRT) class, especially in high-loaded conditions, while maintaining the QoS of real-time (RT) classes. Our scheme with the discarding method can also reduce both the mean waiting time and cell loss ratio of RT classes.

In this letter, we introduce new parameters for classifying digitally modulated unknown QAM and PSK signals. Our two parameters for the classification are the variance of magnitude ratios and the mean of mod 2π phase differences. The gain adjustments of amplitudes are not required for the classification. Five different types of QAM constellations and three different types of PSK constellations are tested and the characteristics of our classification parameters are investigated in various SNR environments. Simulation results demonstrate the effectiveness of our proposed technique.

The goal of this paper is to describe a practical and efficient algorithm for computing in the Jacobian of a large class of algebraic curves over a finite field. For elliptic and hyperelliptic curves, there exists an algorithm for performing Jacobian group arithmetic in O(g2) operations in the base field, where g is the genus of a curve. The main problem in this paper is whether there exists a method to perform the arithmetic in more general curves. Galbraith, Paulus, and Smart proposed an algorithm to complete the arithmetic in O(g2) operations in the base field for the so-called superelliptic curves. We generalize the algorithm to the class of Cab curves, which includes superelliptic curves as a special case. Furthermore, in the case of Cab curves, we show that the proposed algorithm is not just general but more efficient than the previous algorithm as a parameter a in Cab curves grows large.

Semistructured data has no a-priori schema information, which causes some problems such as inefficient storage and query execution. To cope with such problems, extracting schema information from semistructured data has been an important issue. However, in most cases optimal schema information cannot be extracted efficiently, and few efficient approximation algorithms have been proposed. In this paper, we consider an approximation algorithm for extracting "typical" classes from semistructured data. Intuitively, a class C is said to be typical if the structure of C is "similar" to those of "many" objects. We present the following results. First, we prove that the problem of deciding if a typical class can be extracted from given semistructured data is NP-complete. Second, we present an approximation algorithm for extracting typical classes from given semistructured data, and show a sufficient condition for the approximation algorithm to run in polynomial time. Finally, by using extracted classes obtained by the approximation algorithm, we propose a polynomial-time algorithm for constructing a set R of classes such that R covers all the objects to form a database schema.

Can we find in polynomial time an elliptic curve of a given order over a finite field? This paper is concerned with this question which is open since 1986. Consider the partial multivalued function that outputs such an elliptic curve. We characterize the difficulty of computing this function, and show that the polynomial time hierarchy collapses if sat reduces to this function with respect to the polynomial time Turing reducibility, where sat is the partial multivalued function that on input a Boolean formula, outputs a satisfying assignment. We also give a problem that is equivalent to the open question under the Extended Riemann Hypothesis.

We propose an adaptive complex-amplitude texture classifier that takes into consideration height as well as reflection statistics of interferometric synthetic aperture radar (SAR) images. The classifier utilizes the phase information to segment the images. The system consists of a two-stage preprocessor and a complex-valued SOFM. The preprocessor extracts a complex-valued feature vectors corresponding to height and reflectance statistics of blocks in the image. The following SOFM generates a set of templates (references) adaptively and classifies a block into one of the classes represented by the templates. Experiment demonstrates that the system segments an interferometric SAR image successfully into a lake, a mountain, and so on. The performance is better than that of a conventional system dealing only with the amplitude information.

A minimum classification error formulation based on genetic algorithm is proposed for discriminative training of the bigram language model. Results of Chinese dialect identification were reported which demonstrate performance improvement with use of the genetic algorithm over the generalized probabilistic descent algorithm.

With the continuing growth of the World Wide Web (WWW) services over the Internet, the demands for rapid image transmission over a network link of limited bandwidth and economical image storage of a large image database are increasing rapidly. In this paper, a classified binary-tree-structured Self-Organizing Feature Map neural network is proposed to design image vector codebooks for quantizing images. Simulations show that the algorithm not only produces codebooks with lower distortion than the well-known CVQ algorithm but also can minimize the edge degradation. Because the adjacent codewords in the proposed algorithm are updated concurrently, the codewords in the obtained codebooks tend to be ordered according to their mutual similarity which means more compression can be achieved with this algorithm. It should also be noticed that the obtained codebook is particularly well suited for progressive image transmission because it always forms a binary tree in the input space.

In this article, an efficient vector quantization (VQ) scheme called side-match finite-state vector quantization with adaptive block classification is presented for image compression. It makes use of edge information contained in image in additional to the average values of blocks forming the image. In order to achieve low bit rate coding while preserving good quality images, neighboring blocks are utilized to predict the class of current block. Image blocks are mainly classified as edge blocks and non-edge blocks in this coding scheme. To improve the coding efficiency, edge blocks and non-edge blocks are further reclassified into different classes, respectively. Moreover, the number of bits for encoding an image is greatly reduced by foretelling the class of input block and applying small state codebook in corresponding class. The improvement of the proposed coding scheme is attractive as compared with other VQ techniques.

This paper describes a method for training a pattern classifier that will perform well after it has been adapted to changes in input conditions. Considering the adaptation methods which are based on the transformation of classifier parameters, we formulate the problem of optimizing classifiers, and propose a method for training them. In the proposed training method, the classifier is trained while the adaptation is being carried out. The objective function for the training is given based on the recognition performance obtained by the adapted classifier. The utility of the proposed training method is demonstrated by experiments in a five-class Japanese vowel pattern recognition task with speaker adaptation.