This PDF file contains the front matter associated with SPIE Proceedings Volume 6576, including the Title Page, Copyright information, Table of Contents, Introduction (if any), and the Conference Committee listing.

We introduce independent vector analysis (IVA) which is an extension of independent component analysis (ICA) to multivariate components. In a set of ICA mixtures, IVA groups dependent source components across different ICA mixtures and regard them as a multivariate source. This new formulation is an efficient framework for solving the permutation problem in frequency-domain blind source separation (BSS) and its application to n×n speech separation problem has been very successful. In this paper, we present a short tutorial on IVA and summarize the various models that have been proposed to model the frequency components of speech.

We presented a method to reduce noise in signals using a higher-order, correlation-based approach. This paper examines the differences between hard and soft thresholds using the higher-order method, and the use of different wavelets in the denoising algorithm. Using a detection algorithm derived from third-order statistics, we determined if a wavelet coefficient was either mostly noise or mostly signal based on third-order statistics. We found that hard thresholding worked best when compared to soft thresholding but there is the possibility of improvement using soft thresholding.

Support vector machines (SVM) present very interesting features in the field of image processing, but the intensive calculation needed complicates its use in real-time applications. We present an architecture for a SVM which simplifies most of the calculus using an optoelectronic matrix-vector multiplier (OMVM).

In this paper, we present an unsupervised learning with mini free energy for early breast cancer detection. Although an early malignant tumor must be small in size, the abnormal cells reveal themselves physiologically by emitting spontaneously thermal radiation due to the rapid cell growth, the so-called angiogenesis effect. This forms the underlying principle of Thermal Infrared (TIR) imaging in breast cancer study. Thermal breast scanning has been employed for a number of years, which however is limited to a single infrared band. In this research, we deploy two satellite-grade dual-color (at middle wavelength IR (3 - 5&mgr;m) and long wavelength IR (8 - 12&mgr;m)) IR imaging cameras equipped with smart subpixel automatic target detection algorithms. According to physics, the radiation of high/low temperature bodies will shift toward a shorter/longer IR wavelength band. Thus, the measured vector data x per pixel can be used to invert the matrix-vector equation x=As pixel-by-pixel independently, known as a single pixel blind sources separation (BSS). We impose the universal constraint of equilibrium physics governing the blackbody Planck radiation distribution, i.e., the minimum Helmholtz free energy, H = E - T_oS. To stabilize the solution of Lagrange constrained neural network (LCNN) proposed by Szu et al., we incorporate the second order approximation of free energy, which corresponds to the second order constraint in the method of multipliers. For the subpixel target, we assume the constant ground state energy E_o can be determined by those normal neighborhood tissue, and then the excited state can be computed by means of Taylor series expansion in terms of the pixel I/O data. We propose an adaptive method to determine the neighborhood to find the free energy locally. The proposed methods enhance both the sensitivity and the accuracy of traditional breast cancer diagnosis techniques. It can be used as a first line supplement to traditional mammography to reduce the unwanted X-rays during the chemotherapy recovery. More important, the single pixel BSS method renders information on the tumor stage and tumor degree during the recovery process, which is not available using the popular independent component analysis (ICA) techniques.

Statistical Mechanics, which is due primarily to Maxwell, Gibbs, and Boltzmann in the ninetieth century, has proven to be useful model for drawing inferences about the collective behavior of individual objects that interact according to a known force law (which for more general usage is referred to as interacting units.). Collective behavior is determined not by computing F = ma for each interacting unit because the problem is mathematically intractable. Instead, one computes the partition function for the collection of interacting units and predicts statistical behavior from the partition function. Statistical mechanics was united with Bayesian inference by Jaynes [4]. As a continuation, Shannon [7] demonstrated that the partition function assignment of probabilities via the interaction Hamiltonian is the solution to Bayesian assignment of probabilities (based on the maximum entropy method with known means and standard deviations). Once this technique has been applied to a variety of problems and obtained a solution, one can, of course, solve the inverse problem of to determine the solution to an inverse problem to determine what interaction model gives rise to a given probability assignment [1] and [8]. The usage of statistical mechanics allows one can draw general inferences about any complex system including networks [5] by defining "energy", "heat capacity", "temperature", and other thermodynamic characteristics of most complex systems based on the common standard of the Helmholtz free energy. Principle has noted that the aspect of entropy used in reasoning with uncertainty may not be the most appropriate entropy for learning mechanisms [6]. Instead he has explored using Renyi entropy and derived a form of information learning dynamics that has some promising features [2]. To fully realize the potential of the usage of a more generalized entropy to the three aspects of survival, we suggest some connections to the free energy and learning. We also connect some aspects of sensing to probability distributions that suggest why certain search strategies perform better than others. In making these connections, we suggest a fundamental connection waits to be discovered between inference, learning, and related to the manner in which sensing mechanisms perform.

Following our first design concept paper, we have further explored the potential of detection at both the Middle Infrared (Mid-IR) and Long Infrared (Long-IR) spectrum emitted through elevated growth stress of lower GI tract tumors by insertion of a cylindrical fiber-optic lens into the rectum and colon. Electrophysiology suggests that we study the electrical properties of both biological cells and tissues. One may do this by intracellular or extracellular recording. To effectively access the relationship of the sigmoid colon and rectum in humans, it is important to study the electrical and mechanical activation (pressure); we do this by close examination of pacesetter and action potentials, detection of various lower GI tract arrhythmias and studying early developmental symptoms of the "angiogenesis effect". Making use of non-sequential ray tracing software, we seek to design a plastic lens of appropriate index of refraction for the focusing of Long-IR (8-12&mgr;m) onto the axis and a Mid-IR (3-5&mgr;m) on one-half the radius of the cylindrical surface. Intensity (passive dual-color IR spectrograms) will be measured via our minimally invasive device, the so-called Rectumoscope^TM, and correlated with various transcutaneous and invasive electrophysiological measurements.

As a complement to model-based approaches for the analysis of functional magnetic resonance imaging (fMRI) data, methods of exploratory analysis offer interesting options. While unsupervised clustering techniques can be employed for the extraction of signal patterns and segmentation purposes, topographic mapping techniques such as the Self-Organizing Map (SOM) and the Topographic Mapping for Proximity Data (TMP) provide additionally a structured representation of the data. In this contribution we investigate the applicability of two recently proposed variants of these algorithms which make use of concepts from non-Euclidean geometry for the analysis of fMRI data. Compared to standard methods, both approaches provide more freedom for the representation of complex relationships in low-dimensional mappings while they offer a convenient interface for the visualization and exploration of high-dimensional data sets. Based on data from fMRI experiments, the application of these techniques is discussed and the results are quantitatively evaluated by means of ROC statistics.

A methodology for segmentation of multi-component signals buried in additive white Gaussian noise using singular value decomposition (SVD) in the time-frequency domain is proposed. The segmentation problem is posed as a binary statistical hypothesis testing problem. Using the Generalized Likelihood Ratio (GLR), the optimal test statistic is shown to be the sum of squares of the norms of the principal components of the signal in the time-frequency domain. The signal-to-noise ratio (SNR) at the dominant signal frequencies is assumed to be sufficiently high to determine the bandwidth of the signal components. The proposed segmentation methodology is evaluated on phonocardiogram (PCG) signals.

In recent years, dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) has become a powerful complement to X-ray based mammography in breast cancer diagnosis and monitoring. In DCE-MRI the time related development of the signal intensity after the administration of contrast agent can provide valuable information about tissue characteristics at pixel level. The integration of this information constitutes an important step in the analysis of DCE-MRI data. In this contribution we investigate the applicability of three different approaches from the field of independent component analysis (ICA) for feature extraction and image fusion in the context of DCE-MRI data. Next to FastICA, Tree-Dependent Component Analysis and Topographic ICA are applied to twelve clinical cases from breast cancer research with a histopathologically confirmed diagnosis. The outcome of all algorithms is quantitatively evaluated by means of Receiver Operating Characteristics (ROC) statistics. Additionally, the estimated components are discussed exemplarily and the corresponding data is visualized. The study suggests that all of the employed algorithms show some potential for the purposes of lesion detection and subclassification and are rather robust with respect to their parameterization. However, with respect to ROC analysis Tree-Dependent Component Analysis tends to outperform all other algorithms as well as with regarding to the consistency of the results.

It is known that spectral-spatial ICA basis functions of visible color images are similar to some processing elements in the human visual systems in that they resemble Gabor filters and show color opponencies. In this research we study combined spectral-spatial ICA basis functions of multispectral MWIR images. These ICA spectral-spatial basis functions are then used as filters to extract features from multispectral MWIR images. It is hypothesized that learning the added dimension of spectral information along with spatial characteristics of basis functions using ICA improves classification performance for multispectral MWIR images. The images are captured in the 3.0 - 5.0um, 3.7 - 4.2um and 4.0 - 4.5um bands using a multispectral MWIR camera. The phase relationship between the basis functions indicate how the extracted features from the different spectral band images can be combined. We use classification performance to compare features obtained by filtering using multispectral ICA basis functions, multispectral PCA basis functions and opponent Gabor filters.

The usual paradigm for signal processing is to model a signal as a bandlimited function and capture the signal by means of its time samples. The Shannon-Nyquist theory says that the sampling rate needs to be at least twice the bandwidth. For broadbanded signals, such high sampling rates may be impossible to implement in circuitry. Compressed Sensing is a new area of signal processing whose aim is to circumvent this dilemma by sampling signals closer to their information rate instead of their bandwidth. Rather than model the signal as bandlimited, Compressed Sensing, assumes the signal can be represented or approximated by a few suitably chosen terms from a basis expansion of the signal. It also enlarges the concept of sample to include the application of any linear functional applied to the signal. In this paper, we shall give a brief introduction to compressed sensing that centers on the effectiveness and implementation of random sampling.

We extend the 2^nd Gen Discrete Wavelet Transform (DWT) of Swelden to the Next Generations (NG) Digital Wavelet Transform (DWT) preserving the statistical salient features. The lossless NG DWT accomplishes the data compression of "wellness baseline profiles (WBP)" of aging population at homes. For medical monitoring system at home fronts we translate the military experience to dual usage of veterans & civilian alike with the following three requirements: (i) Data Compression: The necessary down sampling reduces the immense amount of data of individual WBP from hours to days and to weeks for primary caretakers in terms of moments, e.g. mean value, variance, etc., without the artifacts caused by FFT arbitrary windowing. (ii) Lossless: our new NG_DWT must preserve the original data sets. (iii) Phase Transition: NG_DWT must capture the critical phase transition of the wellness toward the sickness with simultaneous display of local statistical moments. According to the Nyquist sampling theory, assuming a band-limited wellness physiology, we must sample the WBP at least twice per day since it is changing diurnally and seasonally. Since NG_DWT, like the 2^nd Gen, is lossless, we can reconstruct the original time series for the physicians' second looks. This technique of NG_DWT can also help stock market day-traders monitoring the volatility of multiple portfolios without artificial horizon artifacts.

Video watermarking enforces property right for digital video: a mark is transparently embedded into original data. The true owner is identified by detecting this mark. The robust watermarking techniques allow the mark detection even when the protected video is attacked. Generally, the better the transparency and robustness, the smaller the mark size. We evaluate the maximum theoretical quantity of information which can be inserted into the 2D-DWT coefficient hierarchy, for prescribed transparency and robustness constraints. In order to ensure the accuracy in capacity evaluation, our paper do not relay on any assumption concerning the noise model. Instead, it carries out an in-depth analysis on the statistical behaviour of the real life attacks (StirMark, Gaussian filtering, sharpening, rotation). The experiments are performed on 10 low rate video sequences of 30 minutes each and compares among them three types of bi-orthogonal DWT, namely the (2,2), (4,4), and (9,7). The overall results (theoretical and experimental) are discussed not only for conventional watermarking applications, but for hidden channel, indexing and retrieval applications, as well.

Classification and characterization of text is of ever growing importance in defense and national security. The text classification task is an instance of classification using sparse features residing in a high dimensional feature space. Two standard (of a wide selection of available) algorithms for this task are the naive Bayes classifier and the Rocchio linear classifier. Naive Bayes classifiers are widely applied; the Rocchio algorithm is primarily used in document classification and information retrieval. Both these classifiers are popular because of their simplicity and ease of application, computational speed and reasonable performance. One aspect of the Rocchio approach, inherited from its information retrieval origin, is that it explicitly uses both positive and negative models. Parameters have been introduced which make it adaptive to the particulars of the corpora of interest and thereby improve its performance. The ideas inherent in these classifiers and in second generation wavelets can be recombined into new algorithms for classification. An example is a classifier using second generation wavelet-like functions for class probes that mimic the Rocchio positive template - negative template approach.

We used measures based on entropy to evaluate a method designed to fuse imagery from different sensor types. The method uses different forward transforms of input images and a common transform to reconstruct the final result. We attempted to examine the link between the error in a reconstructed result and its associated entropy.

In this paper, we will discuss a technique of texture image classification using a wavelet decomposition with selective wavelet packet node decomposition. This new approach uses a two-channel wavelet decomposition which is extended to two dimensions. Using the strength as a metric, selective wavelet decomposition is controlled. The metric is used to allow further decomposition or to terminate recursive decompositions. Decision of continuing further decompositions is based on each subband's strength with respect to the strengths of other subbands of the same wavelet decomposition level. Once the decompositions stop, the structure of the packet is stored in a data structure. Using the information from the data structure, dominating channels are extracted. These are defined as paths from the root of the packet to the leaf with the highest strengths. The list of dominating channels are used to train a learning vector quantization neural network.

Many improvements of wavelet-based restoration techniques suggest the use of the total variation (TV) algorithm. The concept of combining wavelet and total variation methods seems effective but the reasons for the success of this combination have been so far poorly understood. We propose a variation of the total variation method designed to avoid artifacts such as oil painting effects and is more suited than the standard TV techniques to be implemented with wavelet-based estimates. We then illustrate the effectiveness of this new TV-based method using some of the latest wavelet transforms such as contourlets and shearlets.

Carbon nanotubes (CNTs) have a potential to be efficient infrared (IR) detection material due to their unique electronic properties. As a one-dimensional nano-structural material, the ballistic electronic transport property makes the noise equivalent temperature difference smaller compared with other semi-conducting materials. In order to verify this unique property, a single pixel CNT-based infrared photodetector is fabricated by depositing the CNTs on the substrate surface and then aligning them to bridge the electrode gap using the atomic force microscopy (AFM)-based nano-robotic system. The photon-generated electron-hole pairs within the carbon nanotube are separated by an external electric field between the two electrodes. The separated carriers contribute to the current flowing through the carbon nanotube and form the photocurrent. By monitoring the photocurrent, the incident infrared can be detected and quantified. Experimental results show the good sensitivity of CNTs to the infrared light.

Nanomanufacturing and nanometrology are key enablers for fulfilling the promise of nanotechnology. Nanomanufacturing (NM) capitalizes on the special material properties and processing capabilities at the nanoscale, and promotes integration of nanostructures to functional micro devices and meso/macroscale architectures and systems, as well as the interfacing issues across dimensional scales. Nanometrology provides the means to measure and characterize process and product performance and covers an expanse of topics including instrumentation, measurement approaches for off-line and in-process production applications, and standards. To meet the needs of this emerging manufacturing community it is important that research on scale-up of nanotechnology for high rate production, reliability, robustness, yield, efficiency and cost issues for manufacturing products and services be pursued. To achieve this, new research directions must include a systems approach that encompasses nanoscale materials and structures, fabrication and integration processes, production equipment and characterization of instrumentation, theory/ modeling/ simulation and control tools, biomimetic design and integration of multi-scale functional systems, three dimensional nanoscale metrology, production-hardened metrology, and other areas driven by industrial applications. Impact is expected in energy systems, electronics, medical/health, transportation, pharmaceutical, chemicals and defense sectors.

Current health care system serving 78M aging baby-boomers is no longer sustainable, as the cost about 1/5 GDP will reach 1/4 GDT when all is retired in decades, unless the system is changed. We design a high-tech safe net to enhance the timeliness of early correct treatment execution (otherwise, causing 1/4 mortality associated with an escalating legal fee waste). We follow the common sense that "a stitch in time saves nine," and adopt the military surveillance know-how in designing early warning health management system, comprising of smart sensor pairs for point-care surveillance. However, the grand plan of affordable smart sensors hardware for households requires an ODM & OEM teaming to conduct parallel designing and sequential marketing strategy. The military software strategy combating a treacherous adversary enemy match well with point cares surveillance overcoming real world microorganism variability. Moreover, such smart military software provides self-reference change detection, not by traditional cohort ensemble average, but by individual own higher order statistics (HOS) independent component analysis (ICA), which take the advantage of known initial condition for each individual and desirable over-sampling daily dynamics. The triggering of warning follows the military algorithms comprising of Receiver Operation Characteristics (ROC) and Automatic Target Recognition (ATR). To further reduce the unwanted false negative rate, a benchmarked is made against the traditional cohort-ensemble baseline average & the upper & lower bounds of variance as adopted by the gatekeepers - Medical Doctors (MD) and Nurses.

The existing methods of data analysis, either the probability theory or the spectral analysis, are all developed by mathematicians or based on their rigorous mathematical rules. For data analysis, we have to face the reality of nonstationarity and nonlinearity in the data. Hilbert-Huang Transform (HHT) is the designated name for the result of Empirical Mode Decomposition (EMD) and the Hilbert Spectral Analysis (HSA) methods, which were both introduced recently by Huang et al. (1996, 1998, 1999 and 2003). The method is adaptive, and specifically for analyzing data from nonlinear and nonstationary processes.

Crack evaluation is essential for effective classification of pavement cracks. Digital images of pavement cracks have been analyzed using techniques such as fuzzy set theory and neural networks. Bidimensional empirical mode decomposition (BEMD), a new image analysis method recently developed, can potentially be used for pavement crack evaluation. BEMD is an extension of the empirical mode decomposition (EMD), which can decompose non-linear and non-stationary signals into basis functions called intrinsic mode functions (IMF). IMFs are monocomponent functions that have well defined instantaneous frequencies. EMD is a sifting process that is non-parametric and data-driven; it does not depend on an a priori basis set. It is able to remove noise from signals without complicated convolution processes. BEMD decomposes an image into two-dimensional IMFs. The present paper explores pavement crack detection using BEMD together with the Sobel edge detector. A number of images are filtered with BEMD to remove noise, and the residual image analyzed with the Sobel edge detector for crack detection. The results are compared with results from the Canny edge detector, which uses a Gaussian filter for image smoothing before performing edge detection. The objective is to qualitatively explore how well BEMD is able to smooth an image for more effective and speedier edge detection with the Sobel method.

We develop an accumulatively effective and affordable set of smart pair devices to save the exuberant expenditure for the healthcare of aging population, which will not be sustainable when all the post-war baby boomers retire (78 millions will cost 1/5~1/4 GDP in US alone). To design an accessible test-bed for distributed points of homecare, we choose two exemplars of the set to demonstrate the possibility of translation of modern military and clinical know-how, because two exemplars share identically the noninvasive algorithm adapted to the Smart Sensor-pairs for the real world persistent surveillance. Currently, the standard diagnoses for malignant tumors and diabetes disorders are blood serum tests, X-ray CAT scan, and biopsy used sometime in the physical checkup by physicians as cohort-average wellness baselines. The loss of the quality of life in making second careers productive may be caused by the missing of timeliness for correct diagnoses and easier treatments, which contributes to the one quarter of human errors generating the lawsuits against physicians and hospitals, which further escalates the insurance cost and wasteful healthcare expenditure. Such a vicious cycle should be entirely eliminated by building an "individual diagnostic aids (IDA)," similar to the trend of personalized drug, developed from daily noninvasive intelligent databases of the "wellness baseline profiling (WBP)". Since our physiology state undulates diurnally, the Nyquist anti-aliasing theory dictates a minimum twice-a-day sampling of the WBP for the IDA, which must be made affordable by means of noninvasive, unsupervised and unbiased methodology at the convenience of homes. Thus, a pair of military infrared (IR) spectral cameras has been demonstrated for the noninvasive spectrogram ratio test of the spontaneously emitted thermal radiation from a normal human body at 37°C temperature. This invisible self-emission spreads from 3 microns to 12 microns of the radiation wavelengths. This pair of cameras are used in the military satellite surveillance imaging operated at 3~5 mid IR band and 8~12 long IR band accompanied by several other UV and visible bands cameras. We can thereby measure accurately both the thermal emission bands at the mid IR and the long IR (X₁ X₂). The spectral ratio will be independent of the depth and imaging environment. Similarly, we will take six times per pair saliva samples (X₁ X₂) inside the upper jaw for three meals daily, of which the dynamics is shown as a delayed mirror image of "blood glucose level". And for which we must design a portable lab "system on chip (SOC)," and the micro-fluidity of pair channels per chemical reactions. According to the same biochemical principle of spontaneity, we apply the identical algorithm to determine both the ratio of hidden malignant and benign heat sources (s₁, s₂) and the blood glucose & other sources (s₁, s₂) leaking into the saliva. This is possible because of the Gibbs isothermal spontaneous process, in which the Helmholtz free energy must be minimized for the spontaneous thermal radiation from unknown mixing of malign and benign sources or the diffusion mixing of glucose s₁ * and other sources s₂ *. We have derived a general formula relating two equilibrium values, before and after, in order to design our devices. Daily tracking the spectrogram ratio and saliva glucose levels are, nevertheless, needed for a reliable prediction of individual malignant angiogenesis and blood glucose level in real time.

SilentGuard delivers substantially new and innovative capability. SilentGuard is a small, unattended, through-the-wall, easy-to-use, undetectable personnel sensor. SilentGuard addresses the need for unattended sensors that can detect personnel through barriers. When configured for a stand alone portable sensor and placed in a passage or within a cleared building, it will alert from a distance to personnel activity within the passage or building by sending a coded alert. When configured as a portable perimeter security sensor, it will provide early warning of personnel intrusion. SilentGuard is a sense-through-wall technology.

Design of current DSP applications using state-of-the art multi-million gates devices requires a broad foundation of engineering skills ranging from knowledge of hardware-efficient DSP algorithms to CAD design tools. The requirement of short time-to-market, however, requires to replace the traditional HDL based designs by a MatLab/Simulink-based design flow. This not only allows the over 1 million MatLab users to design with FPGAs but also to by-pass the hardware design engineer and leads therefore to shorter development time. We have evaluated the Altera/Simulink tool flow used for a University design environment and present design experience of a semester course at FAMU-FSU College of Engineering. We discuss the required background knowledge, key target smart firmware for FPGAs and possible advanced designs, e.g. FFT and multirate filter banks and wavelets designed by students with only basic logic experience.

The design of an microprocessor is a long, tedious, and error-prone task consisting of typically three design phases: architecture exploration, software design (assembler, linker, loader, profiler), architecture implementation (RTL generation for FPGA or cell-based ASIC) and verification. The Language for instruction-set architectures (LISA) allows to model a microprocessor not only from instruction-set but also from architecture description including pipelining behavior that allows a design and development tool consistency over all levels of the design. To explore the capability of the LISA processor design platform a.k.a. CoWare Processor Designer we present in this paper three microprocessor designs that implement a 8/8 wavelet transform processor that is typically used in today's FBI fingerprint compression scheme. We have designed a 3 stage pipelined 16 bit RISC processor (NanoBlaze). Although RISC &mgr;Ps are usually considered "fast" processors due to design concept like constant instruction word size, deep pipelines and many general purpose registers, it turns out that DSP operations consume essential processing time in a RISC processor. In a second step we have used design principles from programmable digital signal processor (PDSP) to improve the throughput of the DWT processor. A multiply-accumulate operation along with indirect addressing operation were the key to achieve higher throughput. A further improvement is possible with today's FPGA technology. Today's FPGAs offer a large number of embedded array multipliers and it is now feasible to design a "true" vector processor (TVP). A multiplication of two vectors can be done in just one clock cycle with our TVP, a complete scalar product in two clock cycles. Code profiling and Xilinx FPGA ISE synthesis results are provided that demonstrate the essential improvement that a TVP has compared with traditional RISC or PDSP designs.

In this paper, we describe a novel algorithm based on diffraction tomography for 3D map generation using the received backscattered radar Electro-Magnetic (EM) field from different spatially distributed multi-static radar sensors. Each sensor at a given time will transmit a radar waveform and the other sensors including the one transmitted will receive the waveform that is backscattered from the objects. A data cube of received data will be created at each sensor by changing the location of sensors. This data cube is used in generating the 3D object profiles at each sensor and then the fused 3D map will be outputted which will contain the fused 3D object profiles or structure obtained from each sensor. If there are more than one object in the field of interest there would be inter object backscattering. This would result in receiving mixed signals. This mixed signal might cause problems in the generation of the 3D map/structure. So to reduce the effect of inter object backscattering we use the probabilistic based blind source separation (BSS) technique for convolutive mixture separation. Before applying the mixture separation technique, we estimate the number of sources. For this we have developed a technique. In this paper, all these techniques are described and also results using real radar backscattered data are provided. A description of how this 3D maps can be used for biomimetics is also provided.

This paper presents an analog processor design for ion sensitive field effect transistor (ISFET)-based flow through system and its application in smart living space. The dynamic flow-cell measurement explores more information compared to stationary measurement and is useful in environmental monitoring and electronic tongue systems. The multi-channel floating source readout circuitry has been developed for flow-through analysis of ion sensitive field effect transistor based array. The flow injection analysis system with two different ISFET structures has been investigated by using performance parameters such as sensitivity, uniformity, response time of pH sensing. In addition, a self-tuning multi-sensor water quality monitoring system based on adaptive-network-based fuzzy interference system (ANFIS) learning method is developed. The results can be directly used in drinking water and swimming pool monitoring for improving living space and quality.

Online aerial vehicle trajectory design and reshaping are crucial for a class of autonomous aerial vehicles such as reusable launch vehicles in order to achieve flexibility in real-time flying operations. An aerial vehicle is modeled as a nonlinear multi-input-multi-output (MIMO) system. The inputs include the control parameters and current system states that include velocity and position coordinates of the vehicle. The outputs are the new system states. An ideal trajectory control design system generates a series of control commands to achieve a desired trajectory under various disturbances and vehicle model uncertainties including aerodynamic perturbations caused by geometric damage to the vehicle. Conventional approaches suffer from the nonlinearity of the MIMO system, and the high-dimensionality of the system state space. In this paper, we apply a Neural Dynamic Optimization (NDO) based approach to overcome these difficulties. The core of an NDO model is a multilayer perceptron (MLP) neural network, which generates the control parameters online. The inputs of the MLP are the time-variant states of the MIMO systems. The outputs of the MLP and the control parameters will be used by the MIMO to generate new system states. By such a formulation, an NDO model approximates the time-varying optimal feedback solution.

In this paper we present the scheme for a control circuit used in an image processing system which is to be implemented in a neural network which has a high level of connectivity and reconfiguration of neurons for integration and trigger based on the Address-Event Representation. This scheme will be employed as a pre-processing stage for a vision system which employs as its core processing an Optical Broadcast Neural Network (OBNN). [Optical Engineering letters 42 (9), 2488(2003)]. The proposed vision system allows the possibility to introduce patterns from any acquisition system of images, for posterior processing.

In the paper we show that the biologically motivated conception of the use of time-pulse encoding gives the row of advantages (single methodological basis, universality, simplicity of tuning, training and programming et al) at creation and designing of sensor systems with parallel input-output and processing, 2D-structures of hybrid and neuro-fuzzy neurocomputers of next generations. We show principles of construction of programmable relational optoelectronic time-pulse coded processors, continuous logic, order logic and temporal waves processes, that lie in basis of the creation. We consider structure that executes extraction of analog signal of the set grade (order), sorting of analog and time-pulse coded variables. We offer optoelectronic realization of such base relational elements of order logic, which consists of time-pulse coded phototransformers (pulse-width and pulse-phase modulators) with direct and complementary outputs, sorting network on logical elements and programmable commutations blocks. We make estimations of basic technical parameters of such base devices and processors on their basis by simulation and experimental research: power of optical input signals - 0.200-20 &mgr;W, processing time - microseconds, supply voltage - 1.5-10 V, consumption power - hundreds of microwatts per element, extended functional possibilities, training possibilities. We discuss some aspects of possible rules and principles of training and programmable tuning on the required function, relational operation and realization of hardware blocks for modifications of such processors. We show as on the basis of such quasiuniversal hardware simple block and flexible programmable tuning it is possible to create sorting machines, neural networks and hybrid data-processing systems with the untraditional numerical systems and pictures operands.

An analysis of the impact of cost function on classifier design is presented. The well know asymptotic probabilistic approach, that invokes the law of large numbers, is extended by incorporating a piece-wise weighted quadratic approximation. This allows different cost functions to be compared and better quantifies the impact of the cost function on the resulting classifier design. In this paper we show how the choice of several well known cost functions are related to (1) Bayesian optimality, (2) classifier complexity, and (3) the ability to estimate decision boundaries. This work extends previous work that relates classifier design to approximations of the Bayesian posterior probability of class membership (e.g., "Any Reasonable Cost Function Can be Used for A Posteriori Probability Approximation" by M. Saerens, et al., IEEE Transactions on NN, September 2002). Several cost functions are analyzed in the paper including the Lp norm and the maximum mutual information (MMI) criterion. An interesting example that supports the theoretical analysis is presented. For the example the Lp norm (with p=1.1) was shown to successfully estimate the Bayesian optimal class decision boundary while the MMI and the L₂ criteria did not. In addition, a finite-version of the theory is presented that bridges the gap between asymptotic theory and strictly finite-size training sets.

This paper presents a fully adaptive store-and-forward routing algorithm in wireless mesh networks (WMNs). In this algorithm, each packet is assigned a specific turn scheme based on its source and destination locations. A full routing adaptivity is achieved for every packet propagated in the network. The deadlock and livelock problems are addressed and solved in this paper. A hierarchy routing structure is proposed to achieve deadlock freedom without introducing of virtual channels or buffer pools. The simulation results show that the proposed routing scheme provides a better fault tolerant performance in comparison with the conventional partially adaptive and dimensional ordered routing algorithms. The system throughput is analyzed theoretically and then validated by the simulations.

The UAVs (Unmanned Air Vehicles) have great potentials in different civilian applications, such as oil pipeline surveillance, precision farming, forest fire fighting (yearly), search and rescue, boarder patrol, etc. The related industries of UAVs can create billions of dollars for each year. However, the road block of adopting UAVs is that it is against FAA (Federal Aviation Administration) and ATC (Air Traffic Control) regulations. In this paper, we have reviewed the latest technologies and researches on UAV navigation and obstacle avoidance. We have purposed a system design of Jittering Mosaic Image Processing (JMIP) with stereo vision and optical flow to fulfill the functionalities of autonomous UAVs.

MatLab/Simulink-based design flows are being used by DSP designers to improve time-to-market of FPGA implementations. ¹ Commonly, digital signal processing cores are integrated in an embedded system as coprocessors. Existing CAD tools do not fully address the integration of a DSP coprocessor into an embedded system design. This integration might prove to be time consuming and error prone. It also requires that the DSP designer has an excellent knowledge of embedded systems and computer architecture details. We present a prototyping platform and design flow that allows rapid integration of embedded systems with a wavelet coprocessor. The platform comprises of software and hardware modules that allow a DSP designer a painless integration of a coprocessor with a PowerPC-based embedded system. The platform has a wide range of applications, from industrial to educational environments.

Various kinds of classification methods have been developed. However, most of these classical methods, such as Back-Propagation (BP), Bayesian method, Support Vector Machine(SVM), Self-Organizing Map (SOM) are arrogant. A so-called arrogance, for a human, means that his decision, which even is a mistake, overstates his actual experience. Accordingly, we say that he is a arrogant if he frequently makes arrogant decisions. Likewise, some classical pattern classifiers represent the similar characteristic of arrogance. Given an input feature vector, we say a classifier is arrogant in its classification if its veracity is high yet its experience is low. Typically, for a new sample which is distinguishable from original training samples, traditional classifiers recognize it as one of the known targets. Clearly, arrogance in classification is an undesirable attribute. Conversely, a classifier is non-arrogant in its classification if there is a reasonable balance between its veracity and its experience. Inquisitiveness is, in many ways, the opposite of arrogance. In nature, inquisitiveness is an eagerness for knowledge characterized by the drive to question, to seek a deeper understanding. The human capacity to doubt present beliefs allows us to acquire new experiences and to learn from our mistakes. Within the discrete world of computers, inquisitive pattern recognition is the constructive investigation and exploitation of conflict in information. Thus, we quantify this balance and discuss new techniques that will detect arrogance in a classifier.

Large cortical models based on differential equations may require significant computations to converge, in addition to the computations required to simulate learning. Fortunately, sensitivity analysis for such models can be done using the implicit function theorem (IFT), as shown by McFadden in 1993 for a model with "virtual lateral inhibition" (VLI) in which inhibition is based on competition for activation, rather than on direct reduction of activation levels. The current work reviews recent neurobiological work on the nature of inhibition, and also reports new results on numerical issues that arise in the analysis of VLI models of cortical networks. The IFT technique is at least an order of magnitude faster than numerical ODE solvers. A new explanation for inhibition based on energy resource sharing is proposed.

In this paper a watermarking technique for Intellectual Property Protection (IPP) of FPGA-based systems is proposed. The aim is to protect the author rights of reusable IP cores by means of a digital signature that uniquely identifies both the original design and the design recipient. The proposed watermarking technique relies on a procedure that spreads the digital signature in cells of memory structures at Hardware Description Language (HDL) design level, not increasing the area of the system. This signature is preserved through synthesis, placement and routing processes. The technique includes a procedure for signature extraction requiring minimal modifications to the system. Thus, it is possible to detect the ownership rights without interfering the normal operation of the system and providing high invulnerability. To illustrate the properties of the proposed watermarking technique, both protected and unprotected design examples are compared in terms of area and performance. The analysis of the results shows that the area increase is very low while throughput penalization is almost negligible.

A neural network algorithm is proposed which can conceal different files effectively such as *.exe, *.com, *.doc, *.txt and self-defined file formats. First, the important contents of the file are coded into a binary array. The total number of 0s and 1s is N. Second, to make the neural network learn the sample space, N pixel values and their closely relevant pixel values are randomly chosen from a color BMP format image and used to train the neural network, thus we get a group of network parameters and its outputs Y1. Each element of Y1 is increased by 0 or 1 according to the zeros and ones from the array, the increased Y1is called Y2. Third, using the transmitted parameters, a receiver can restore the neural network. Network outputs Y3(Y1) can also be obtained by simulating the restored neural network with the seed pixel values. Finally, the encrypted information can be decoded by Y2 and Y3. The acquisition of parameters and Y1 is different when the neural network is trained each time, so the algorithm has the characteristic of a one-time pad, which can enhance the correspondence security. Because the network colligates the chosen pixel values and their closely relevant pixel values, a cryptanalyst can not restore the network parameters and Y3 easily. Without the seed picture and the password, he does not know the encrypted data even if he knows the network parameters and Y2. If he only has the seed picture, he does not know the encrypted contents either, because there is no other information in the picture, which just is used to train the network. Using the same algorithm, the fragile watermark can be embedded into Y1 simultaneously. Besides telling you whether Y2 or network parameters have been tampered with, the fragile watermark could as well, reflecting the distortion status in the spatial domain of the tampered image. Therefore, the proposed method is of significance in practice.

A hybrid algorithm of using a wavelet transform and a neural network is presented which solves the problems confronted in public watermarking systems. First, to get the wavelet coefficients, db1 wavelet is used to decompose the selected image. Second, to ensure better quality of the watermarked image, some wavelet coefficients and their closely relevant wavelet coefficients are randomly selected from the wavelet coefficients decomposed by the low pass filter and used to establish the relational model by using a neural network. Third, the bit information of the watermark is also enlarged by increasing the amount of zeros or ones and then one bit of the results is embedded by adjusting the polarity between a chosen wavelet coefficient and the output value of the model. Finally, a new image with watermark information is reconstructed by using the modified wavelet coefficients and other unmodified wavelet coefficients. On the other hand, the process of retrieving the watermark is the inverse of the embedding process. The embedded watermark can also be retrieved by using the hybrid algorithm and the restore function without knowing the original image and watermark. Experimental results show that the proposed technique is very robust against some image processing operations and JPEG lossy compression. Meanwhile, the extracted watermark can be proved by the proposed method. Because of the neural network, the proposed method is also robust against attack of false authentication. Therefore, the hybrid algorithm can be used to protect the copyright of one important image.

To avoid interference and achieve spectral effectiveness one needs to calculate the spectra of the surroundings then form the basis function needed for modulation out of the clear portion of the spectrum. The calculation of the spectra over a short time as required for some applications could be problematic. These problems reveal themselves in the form of broadened spectral picks and tall side lobes. The solution to remove broadening effects and the side lobes proposed in this report and tested for some frequencies is the constrained iterative spectral deconvolution. In this report, we will provide the simulation results that indicate one period of the interfering signal is all that is need to successfully remove all side effects resulting from short time domain sampling of the environment.