Showing papers in "Iete Journal of Research in 2002"

A Survey on Current Content based Image Retrieval Methods

Abstract: Retrieving information from the Web is becoming a common practice for internet users. However, the size and heterogeneity of the Web challenge the effectiveness of classical information retrieval techniques. Content-based retrieval of images and video has become a hot research area. The reason for this is the fact that we need effective and efficient techniques that meet user requirements, to access large volumes of digital images and video data. This paper gives a brief survey of current CBIR (Content Based Image Retrieval) methods and technical achievement in this area. The survey includes a large number of papers covering the research aspects of system design and applications of CBIR, image feature representation and extraction, Multidimensional indexing. Furthermore future research directions are suggested.

Human Face Detection and Tracking using Skin Color Modeling and Connected Component Operators

Abstract: Face Recognition (FR) systems are increasingly gaining more importance. Face detection and tracking in a complex scene forms the first step in building a practical FR system. In this paper, a method to detect and track human faces in color image sequences is described. Skin color classification and morphological segmentation is used to detect face(s) in the first frame. These detected faces are tracked over subsequent frames by using the position of the faces in the first frame as the marker and detecting for skin in the localized region. Specific advantages of this approach are that skin color analysis method is simple and powerful, and the system can be used to detect/track multiple faces.

Evolutionary Computation for the Recognition of On-Line Cursive Handwriting

Abstract: This paper describes a system that recognizes on-line cursive handwriting. The system was specialized on Arabic script, but it may be adapted to work on any other language. A genetic algorithm is used to select the best combination of characters recognized by the hierarchical Beta neuro-fuzzy system. The handwritten words are modeled by a neuro-physiological theory of movement generation predicting that the main features extracted from each character are the parameters of the equation describing the curvilinear velocity of the script. The evolutionary approach proposed here permits the recognition of cursive handwriting with a segmentation procedure allowing overlapped strokes having neuro-physiological meaning. We also present the experimental results obtained when using the system to recognize on-line handwritten Arabic.

Active Camera Networks and Semantic Event Databases for Intelligent Environments

Abstract: In the future, intelligent rooms, with embedded multimodal sensory systems and semantic event databases, will support effective and efficient transactions of human activities and interactions. We are pursuing rigorous experimental investigations towards the development of such intelligent environments. In this paper we describe the overall system specification and general framework guiding our development. We also present details of the modules associated with the control and interpretation of video information acquired by a network of cameras and a novel semantic event database for characterization and recognition of activities. Evaluations of the system in a custom-built intelligent room are also presented.

Developing a Gesture-based Interface

Abstract: A gesture-based interface involves tracking a moving hand across frames, and extracting the semantic interpretation corresponding to the gesture. This is a difficult task, since there is a change in both the position as well as the appearance of the hand. Further, such a system should be robust to the speed at which the gesture is performed. This paper presents a novel attempt at developing a hand gesture-based interface. We propose an on-line predictive EigenTracker for the moving hand. Our tracker can learn the eigenspace on the fly. We propose a new state-based representation scheme for hand gestures, based on the eigenspace reconstruction error. This makes the system independent of the speed of performing the gesture. We use learning for adapting the gesture recognition system to individual requirements. We show results of successful operation of our system even in cases of background clutter and other moving objects.

A Fast Method for Image Mosaicing using Geometric Hashing

Abstract: The general problem of mosaicing is to create a single seamless image by aligning a series of spatially overlapped images. The result is an image with a field of view greater than that of a single image. Traditionally this research has been aimed at stitching together images taken by aerial or satellite reconnaissance equipment. With the advancement of personal computing equipment, the creation of image mosaic has entered the consumer market. Thus, automation of the process is an important issue. This paper proposes a new method for automatic generation of mosaics using Geometric Hashing. This speeds up the matching process. We show the application of our method on two important cases namely, rigid planar motion and panoramic mosaics. We provide experimental results in support of our proposed method.

Anisotropic diffusion in Image processing: Methods based on Physics, statistics and Geometric Curve Evolution

Abstract: Heat diffusion equations in Physics, have yielded promising results in Scale Space vision as well as edge preserved smoothing that are required in image analysis, restoration and segmentation. Here we will be considering, a parabolic class of 2D second order diffusion PDE (partial differential equations) under nonlinear case, which will be used as smoothing filter preserving edge features. We review various methods that have been proposed based on the theory of Physics, Statistical and Geometric Curve Evolution and observe their performance. The thrust is to make an effort in creating awareness on different anisotropic diffusion methods in Digital Image Processing, since it proves to be an efficient early processing tool for Image Analysis. We will also observe a computationally efficient method based on Geometric Curve Evolution proposed by us.

A Scalable Framework for High-Resolution Immersive Displays

Abstract: We introduce an immersive display framework that is scalable, easily re-configurable, and does not constrain the display surface geometry The system achieves very-high resolution display through synchronized rendering and display from multiple PCs and light projectors The projectors can be placed in a loose configuration and calibrated at run time A full display is composed of these underlying display devices by blending overlapping regions and pre-warping imagery to correct for distortions due to display surface shape and the viewer's position The effect is a perceptually correct display of a single high-resolution frame bufferA major contribution of the work is the addition of cameras into the display environment that assist in calibration of projector positions and the automatic recovery of the display surface shape In addition, a straightforward synchronization framework is introduced that facilitates communication between the multiple rendering elements for calibration, tracking the user's view

Speech recognition using Modular General Fuzzy Min-Max Neural Network

Abstract: In this paper, we propose the Modular General Fuzzy, Min-Max Neural Network (MMGFMM). MMGFMM contains number of modules of Modified General Fuzzy Min-Max Neural Network (MGFMM). In MGFMM, the transfer function of output layer of General Fuzzy Min-Max Neural Network (GFMM) is modified. The performance of GFMM, MGFMM and MMGFMM to recognition of spoken Marathi (Language spoken in the state of Maharashtra, India) digits is reported. MMGFMM has shown better average recognition accuracy compared to GFMM by 15.1% in speaker dependent mode and by 11.9% in speaker independent mode.

Fuzzy Logic based Speed Controllers for Vector Controlled Induction Motor Drive

Abstract: This paper presents a comparative study of Proportional Integral (PI), Fuzy Logic (FL), Fuzzy Pre-compensated Proportional Integral (FPPI), Fuzzy Poportional Integral (FPI), and Hybrid (of FL and PI) speed controllers for vector controlled induction motor drive An indirect Vector Control (VC) strategy is employed for the control of curent controlled voltage source inverter (CC-VSI) fed squirrel cage induction motor drive (SCIMD) The simulated response of a CC-VSI fed cage induction motor drive is presented for different modes of operation such as starting, speed reversal, load application and removal to demonstrate the effectiveness of the various mentioned controllers

Cascade Control of Interconnected System Using Neural Network

Abstract: This paper deals with the application of neural networks in the cascade control of an interconnected system. In this scheme, the neural network has been used to develop the primary and secondary controllers. The performance and stability of the neural controller is affected by the selection of the network parameters. In this work an adaptive change of the learning rate and momentum factor is proposed. The performances of the controller for the set point tracking and the disturbance rejection show that an efficient neural control system can be realized by using adaptive learning rate and momentum factor. In order to improve the transient as well as steady state responses, a linear feedback compensator (LFBC) has been added in parallel with the primary and secondary control network. The feasibility of method has been tested on the example of simulated system. The control strategy further experimentally tested on a laboratory set-up of the continuously stirred tank (CST) process. The robustness of the proposed system has been confirmed with respect to its performance for the set point tracking and also under the influence of load disturbance imposed on the system. The performance of the various control strategy has been compared in terms of integral square error (ISE).

Speech Recognition with Emphasis on Wavelet based Feature Extraction

Abstract: In this paper some of the commonly used feature extraction techniques are presented and a new set of features based on the Discrete Wavelet Transform (DWT) and Admissible Wavelet Packet Transform (AWPT) is presented for the recognition of phonemes. These features overcome the problem of shift variance and speaker dependence encountered in the earlier features derived by using wavelet transform. Further study on the earlier proposed energy features derived by DWT is carried out and AWPT is proposed for phoneme recognition to overcome the problems with DWT based features. Further a new set of features based on the logarithmic compression of the energy is proposed which shows considerable improvement in the recognition performance.

Blind Equalization of Nonlinear Communication Channels using Modified FLANN

Abstract: Adaptive blind equalization has gained widespread use in communication systems that operate without training signals. For the nonlinear channels, however, the linear equalizers are not suitable. Nonlinear mapping capability of neural networks makes them a suitable choice for the equalization of nonlinear channels. In this paper, the application of modified Functional Link Artificial Neural Network (FLANN) with adaptable output node alongwith its learning rule for the blind equalization of nonlinear communication channels is presented. This modification in the FLANN helps in achieving faster convergence. The performance of the proposed network is compared with that of Radial Basis Function (RBF) blind equalizer and the linear Constant Modulus Algorithm (CMA). The small size and simple learning rules make this network suitable for high speed blind equalization.

Image Mosaicing from Video Sequences

Abstract: The task of assembling individual frames from a video stream into a single giant sized image is known as image mosaicing. The present scenario deals with a real world situation where the video frames and the digital flight data are obtained from a sensor mounted on an aircraft. The paper describes different approaches of automatic image mosaicing suitable for the present application.The analytical method finds out matching points between the successive image pairs using some sort of similarity measures and use generalized transformation equation to calculate translation, rotation and scaling. The Fourier method uses Fourier transform to calculate the translation and rotation between successive image frames and assemble the images on a large geo-referenced canvas. In both the cases, perspective distortion in the image due to forward-looking sensor is eliminated.The algorithms, implemented in a SGI workstation, are used to generate mosaics of terrain from the stream of video imagery obtained during actual f...

Multiwavelet and Image Compression

Abstract: Advances in wavelet transforms and quantization methods have produced algorithms capable of surpassing the existing image compression standards like the Joint Photographic Experts Group (JPEG) algorithm. For best performance in image compression, wavelet transforms require filters that combine a number of desirable properties, such as orthogonality and symmetry. However, the design possibilities for wavelets are limited because they cannot simultaneously possess all of these desirable properties. The relatively new field of Multiwavelets shows promise in removing some of the limitations of wavelets. Multiwavelets offer more design options and hence can combine all desirable transform features. It also shows that the latest Multiwavelet filters can give performance equal to, or in many cases superior to, the current wavelet filters. The performance of Multiwavelet packets is also explored and is shown to be competitive to that of wavelet packets in some cases.

Study of Channel Assignment Strategies for Handoff and Initial Access in Mobile Communication Networks

Abstract: An efficient allocation of resources is the key to high system capacity. In this paper, we present a new channel assignment strategy for handover and initial access in a cellular communication network. The objective is to improve the network throughput or the probability of call completion. We develop a model besed on a nonpreemptive priority queueing discipline. New calls, which originate within the cell at a Poisson rate, are queued if all channels are occupied. Handover requests arriving at a Poisson rate are also queued if no channel is available at the time of arrival. Handover requests are served first as soon as channels become available. If the handoff queue is empty, waiting new calls are served. The performance criteria of interest are: probability of call completion, probability of handover dropping, probability of new call blocking and carried versus offered traffic. Our scheme is seen to provide better call completion probability or throughput, less call blocking without much affecting the fo...

A Novel Filtering Scheme for Impulse Noise Cancellation

Abstract: This paper proposes a novel filtering scheme to denoise images contaminated by impulse noise of high densities This scheme utilizes two stage operations In the first stage, corrupted image is passed through the Artificial Neural Network (ANN) detector to identify the noisy pixels by considering their surrounding neighborhood The corrupted pixels are passed through the Discrete Wavelet Transform (DWT) based pre-filter and the filtered pixels are replaced back in the image The pre-filtered image is further passed through the ANN detector to capture any residual corruption A novel selective spatial filtering scheme is proposed for the corrupted pixels The proposed scheme is simulated using standard images under different noise conditions The comparative performance study shows the superiority of the proposed scheme over the existing standard filtering schemes both in terms of noise rejection and edge retention capability even under high noise conditions

Thermal characteristics of back cooled impingement cooler for packaging of high power semiconductor laser diode arrays

Abstract: A simple impingement type of water cooled heat sink has been fabricated and characterized for packaging of high power semiconductor laser diode arrays. A single chilled water jet Impinging normally on the back plane of copper submount is used to achieve enhanced heat transfer coefficient for convective cooling. The cooler is fabricated out of copper (OFHC), which consists of a chamber ( 3x9x12 mm 3 ) with one pair of inlet and outlet tubes for chilled water flow at one end and 3 mm thick copper submount at the other end. The thermal characteristics of the cooler is measured as a function of heater power, heater temperature and chilled water pressure of up to 30PSI. The minimum thermal resistance obtained is 0.6°C/Watt for submount surface area of 3.2 x 10mm 2 . Experimental thermal resistance values closely match with theoretical calculations. The heat sink fabrication and characterization are discussed.

Hierarchical Recognition of Dynamic Hand Gestures for Telerobotic Application

Abstract: In this paper a scheme is presented for recognizing hand gestures using the output of a hand tracker which tracks a rectangular window bounding the hand region. A hierarchical scheme for dynamic hand gesture recognition is proposed based on the dominant feature trajectories using an a priori knowledge of the way in which each gesture is performed. A state representation is obtained from the dominant feature trajectories, where the states correspond to perceptually important segments of hand movement characterizing the gesture.

An Algorithm for Designing Optimal Gabor Filter for Segmenting Multi-Textured Images

Abstract: This paper presents a method for the design of single Gabor filter for segmenting multi-textured images. The features are extracted by filtering with a linear filter and estimating the local energy of the filter response. Gabor filters have been applied successfully to the segmentation of textured images. Previous investigators have used bank of filters, where the filter parameters were predetermined and not optimized for particular task. A model of feature extraction process is required for the optimization, which is developed and assessed to get a single Gabor filter.The approach is assessed by supervised segmentation experiments and includes the design of Gabor filter, Gaussian filter, classifier and post processing. The classifier uses minimum distance algorithm and post processing uses morphological operators to remove spurious misclassifications. Results are presented that confirm the efficiency of the post processing method and support underlying mathematical models.

Optimal design of delay-bounded WDM networks using a genetic algorithm

Abstract: We address the problem of real-time delay-bounded multicasting in wavelength-division multiplexed networks to avoid problems of synchronization between video and audio frames. We describe a genetic algorithm based technique to synthesize WDM network topologies that can, with a very high degree of confidence, assure that the multicast traffic is delivered in user-specified limits on time. Unlike existing approaches to WDM network design, we first find a virtual topology that can meet the delay constraints. An embedding of virtual links into physical links is then carried out, followed by an assignment of wavelengths to virtual links. The problem of finding the virtual topology is difficult because of a large number of parameters. A number of heuristic approaches have been proposed to solve such optimization problems. In our approach we aim to explore the suitability of Genetic Algorithms to solve the WDM network design problem. A genetic algorithm can explore a far greater range of potential solutions to a problem than do conventional approaches. The advantage of a genetic algorithm, compared with other algorithms which use a single initial guess e.g. gradient, descent is to use more Information of estimation region, and to decrease the probability of falling into local minimum. We describe quantitative and qualitative results obtained by using our software tool on several benchmark examples.

Edge Detection in Range Images using Nonlinear Laplacian Operators

Abstract: In this paper, application of a nonlinear laplacian operator for edge detection in range images is discussed. The nonlinear laplace operator which had earlier been shown to be effective for edge detection in intensity images with low and medium signal-to-noise ratios is modified to detect the roof edges in range images. Differences in surface normal in a neighborhood around a pixel in the image is used here in contrast to the differences in gray values used in intensity images. The images are smoothened using median filters of suitable size before applying the laplace operator. Results with synthetic and natural range images are presented.

Extending Real Time Change-Detection Techniques to Mosaic Backgrounds and Mobile Camera Sequences in Surveillance Systems

Abstract: This paper shows a method for extending efficient algorithms for scene understanding already developed and tested for fixed cameras to a mobile camera environment. Real-time change detection methods for mobile-head cameras are introduced. The architecture of the system can be divided in two phases. During the off-line phase the system creates a panoramic multi-layer background image using a small number of static background images. In the on-line phase the system compares the acquired images with a portion of the panoramic background. Different approaches to produce the change detection images are analyzed. Experimental results are presented in order to validate the proposed methods; their evaluation is performed by using receiving operator characteristic (ROC) curves. The Neyman-Pearson statistical criterion has been used for selecting of optimal change detection threshold. The presented results, in terms of probabilities of false and correct detection rates and real-time behavior, show that one of the studied methods can be used as the basis for higher level modules of an automatic video-surveillance system.

Evolutionary Modeling with Hierarchy in Innovative Computer-Aided Circuit Design

Abstract: In this article new approaches for innovative computer-aided design using evolutionary modeling are investigated. It is shown that the prominent features of evolutionary modeling in innovative design are elitism and mutation. Dynamic parameters and a dynamic coding of the genotype and phenotype for improving the evolutionary search for computer-aided design are proposed. Some examples of computer-aided circuit design are discussed.

Depolarization of Microwave/millimeter Wave in Sand and Dust Storms

Abstract: A theoretical investigation has been carried out to evaluate the effect of depolarization on the propagation of microwave/millimeter wave in sand and dust storms. The storm is considered to have three main constituents i.e, sand, silt and clay. An equation has been developed for depolarization In which the concept of scattering by dipole as well as back scattering has been utilized for this purpose. It has been found that the depolarization and loss due to depolarization caused by dust particles depend directly on orientation of Incident beam, frequency and particle density. The results obtained are in good agreement with experimental resuits reported eariler. This verifies the accuracy of theory developed.

Texture Element Extraction via Cepstral Filtering in the Radon Domain

Abstract: We present a procedure based on one-dimensional cepstral filters in the Radon domain to extract texture elements or textons from images with (quasi-) periodic or ordered texture. With this approach, no assumption is required on the homogeneity of the texton. By applying the cepstral filter in the Radon domain, the difficulties associated with two-dimensional cepstral analysis and phase unwrapping are obviated. The necessity of a weighting function as a preprocessing step and details of wavelet extraction in the Radon domain are discussed. The method should facilitate structural analysis of ordered texture and the constituent textons.

Computation of steady-state response in power Electronic circuits

Abstract: A new simulator capable of "direct" steady-state time-domain simulation of circuits with periodic inputs is described. A Newton-Raphson iterative procedure is used to compute the state variables such that, after integrating for one period, they return to their original values. Several Power Electronic circuit examples are discussed. It is shown that the direct method is computationally far superior than the conventional method in which transient simulation is carried out until all non-periodic transients have decayed completely. The new simulator is a public-domain program and can be obtained from the authors.

A Heavy Tailed Probability Distribution Function to Model Self-similar Traffic in ATM Networks

Abstract: Recent findings have brought out the fractal or self-similar nature of network traffic. Self-similar processes have so far been defined only in terms of their auto-correlation functions, which are mathematically complex and not very tractable. In this paper, we present a simple distribution function that models self-similar arrivals by capturing its heavy tailed property.

A novel pwm for harmonic reduction and its application to ac-dc converters

Abstract: The paper describes a novel PWM scheme called inverted-sine PWM (ISPWM) which uses a sinusoidal reference and an inverted-sine carrier. The ISPWM, when applied to a rectifier, has a better harmonic elimination and a higher average output voltage compared to a sine PWM (SPWM). The harmonic content of the ISPWM output for different values of the modulation index is computed and compared with that of a SPWM. The complete circuit for generating the ISPWM control signal for single and three phase ac-dc converters is developed. Experimental waveforms of voltages and currents are presented.

Prediction of Engineering Manpower using Neural Network and Genetic Algorithm

Abstract: The smooth working of industry depends on the availability of proper engineering manpower. If properly qualified and experienced technical personnel are not available, the industry cannot run in the most efficient way. Here an effort is made to assess the engineering manpower requirement (personnel belonging to mechanical engineering) in certain industry group (Steel Manufacturing) in the state of West Bengal in India for the next five years. An approach consists of artificial neural network preferably feed forward back propagation with genetic algorithm is tested and used for the assessment of engineering manpower. In this approach the data is initially fuzzified, and the fuzzified data is used to train an artificial neural network. The output of the trained neural network is defuzzified and the operators of genetic algorithm (GA) are applied on the defuzzified data until the average error lies below a particular value. Certain statistical functions i.e. linear, exponential, curvilinear (parabolic) equat...