CHIMAN KWAN

Bayer pattern has been widely used in commercial digital cameras. In NASA's mast camera (Mastcams) onboard the Mars rover Curiosity, Bayer pattern has also been used in capturing the RGB bands. It is well known that debayering, also known as demosaicing in the literature, introduces artifacts such as false colors and zipper edges. In this paper, we first present four fusion approaches, including weighted and the well-known alpha-trimmed mean filtering approaches. Each fusion approach combines demosaicing results from seven debayering algorithms in the literature, which are selected based on their performance mentioned in other survey papers and the availability of open source codes. Second, we present debayering results using two benchmark image data sets: IMAX and Kodak. It was observed that none of the seven algorithms in the literature can yield the best performance in terms of peak signal-to-noise ratio (PSNR), CIELAB score, and subjective evaluation. Although the fusion algorithms are simple, it turns out that the debayering performance can be improved quite dramatically after fusion based on our extensive evaluations. In particular, the average PSNR improvements of the weighted fusion algorithm over the best individual method are 1.1 dB for the IMAX database and 1.8 dB for the Kodak database, respectively. Third, we applied the various algorithms to 36 actual Mastcam images. Subjective evaluation indicates that the fusion algorithms still work well, but not as good as the existing debayering algorithm used by NASA.

WorldView 3 (WV-3) is the first commercially deployed super-spectral, very high-resolution (HR) satellite. However , the resolution of the shortwave infrared (SWIR) bands is much lower than that of the other bands. In this letter, we describe four different approaches, which are combinations of pansharpening and hypersharpening methods, to generate HR SWIR images. Since there are no ground truth HR SWIR images, we also propose a new picture quality predictor to assess hypersharpening performance, without the need for reference images. We describe extensive experiments using actual WV-3 images that demonstrate that some approaches can yield better performance than others, as measured by the proposed blind image quality assessment model of hypersharpened SWIR images.

The Mars Science Laboratory is a robotic rover mission to Mars launched by NASA on November 26, 2011, which successfully landed the Curiosity rover in Gale Crater on August 6, 2012. The Curiosity rover has two mast cameras (Mastcams) that acquire stereo images at a number of different wavelengths. Each camera has nine bands of which six bands are overlapped in the two cameras. These acquired stereo band images at different wavelengths can be fused into a 12-band multispectral image cube, which could be helpful to guide the rover to interesting locations. Since the two Mastcams' fields of view are three times different from each other, in order to fuse the left-and right-camera band images to form a multispectral image cube, there is a need for a precise image alignment of the stereo images with registration errors at the subpixel level. A two-step image alignment approach with a novel utilization of existing image registration algorithms is introduced in this paper and is applied to a set of Mastcam stereo images. The effect of the two-step alignment approach using more than 100 pairs of Mastcam images, selected from over 500000 images in NASA's Planetary Data System database, clearly demonstrated that the fused images can improve pixel clustering and anomaly detection performance. In particular, registration errors in the subpixel level are observed with the applied alignment approach. Moreover, the pixel clustering and anomaly detection performance have been observed to be better when using fused images.

This paper proposes a novel approach to sensor and actuator integrity monitoring in a dynamic system. Multiple sensor and actuator faults can be detected. Furthermore, faulty sensors and actuators are isolated by contribution analysis. Most importantly, fault magnitudes can be correctly estimated and failed sensors or actuators outputs can be reconstructed. The proposed approach is robust to disturbances, minimizes false alarms, while achieving maximized sensitivity to any faults. Numerical examples justify correctness and validity of the developed methodology.

This paper presents a practical approach to target detection for hyperspectral images. In target detection, it is normally assumed that the ground truth target signatures collected in a laboratory are available and one then uses them to search for targets in a given image. However, directly applying the laboratory signatures to the real data is not appropriate due to environmental differences between the ground truth data and real data. Conventional atmospheric compensation schemes such as the use of MODTRAN can help to improve the target detection performance. However, the computational load is huge and thus real-time applications may prohibit this compensation approach. We present results of an alternative compensation technique known as in-scene compensation, which is appealing as no complicated techniques such as MODTRAN are needed. Two in-scene methods for visible near-infra-red/short-wave infrared range have been developed in the literature: empirical line method (ELM) and vegetation normalization (VN). Both approaches have advantages and disadvantages. We propose a hybrid in-scene compensation method that can be considered as a combination of ELM and VN and we call our method ELM augmented VN (EAVN). One key advantage of EAVN is that it combines the advantages of ELM and VN and eliminates their disadvantages. Compared to ELM, there is no need for two or more known target pixels in the test scene. Compared to VN, there is no need for dark pixels. Extensive experimental results using ground-based sensor data showed that the EAVN algorithm provides excellent compensation to environmental changes. After compensation, the receiver operating characteristics performance of target detection has been significantly improved by orders of magnitude in a number of cases, as compared to two standard compensation methods: quick atmospheric correction and internal average relative reflectance correction.

Remote sensing and its applications have gained more and more attention from researchers in recent years. One clear indicator can be seen from the 2016 International Geo-science and Remote Sensing Symposium (IGARSS), which has received over 3,000 papers. This field is expanding and evolving rapidly. The aim of this special issue is an attempt to capture a small section of recent advances in remote sensing. We would like to thank all the contributing authors, reviewers, and journal staffs for making this special issue a reality. As will be seen shortly, the 6 papers indeed cover a wide range of remote sensing applications using airborne and space-borne instruments. The paper by H. Li et al. addressed an important problem in many state-of-the-art multispectral airborne imagers. Due to significant difference in intensity between different band images, image registration becomes very difficult even though the multispectral imager consists of identical monochrome cameras equipped with different bandpass filters. The authors proposed a two-stage image registration algorithm. The first stage is to use phase correlation method to calculate the parameters of a coarse-offset relationship between different band images. The second stage uses the scale invariant feature transform (SIFT) to detect the feature points. Actual experimental data were used to demonstrate the proposed algorithm. It was seen that conventional SIFT-only method failed whereas the proposed method can still achieve good registration performance. Land cover land use (LCLU) classification is important for environment monitoring and urban planning. The paper by J. Jiao and Z. Deng focused on building and tree detection algorithms by using improved superpixels from large high-resolution urban aerial images. The authors also proposed a method to calculate the tree parameters using a cost function and information from shadows. Experiments showed that their method is fast and robust, while still being simple and efficient, and they also indicate that the shadow is a good feature to estimate the tree height. The results of proposed algorithms have great potential for generating 3D urban models. Airborne electromagnetic methods (AEM) systems are important for estimating abundance of natural resources. Typically, inductance is being measured using a rectangular loop onboard an aircraft. One serious problem with existing systems is that the effect of a finite-conducting ground on the inductance of the transmitting loop was neglected, or the ground was handled as a perfect conductor. In other words, there was no accurate method to evaluate ground's effect on the inductance of the transmitting loop. Consequently, the measurement will not be accurate. X. Jia et al. proposed a new and efficient algorithm to calculate ground's effect on the inductance of a rectangular loop. An experiment was constructed afield, showing that the inductance increased gradually when the loop was lifted up from 0 m to 30 m, which supported the algorithm positively. Using satellite data for generating vegetation profiles has the advantage of large area coverage. Due to the presence of clouds in the data, a composite 16-day period is normally used where at least a certain number of cloud free data are present. However, 16-day period vegetation profiles are less sensitive to real-time changes due to the composite period used for the bidirectional reflectance distribution function (BRDF) model. The paper by S. Kim et al. investigated the impact of different composite periods on the vegetation profiles. Geostationary Ocean Color Imager (GOCI) was

Assessment of damages due to fire, drought, flood, land slide, etc., using hyperspectral images from Hype-rion, AVIRIS or HyspIRI has challenging issues. The effects of different illumination, atmospheric conditions and varying sensor/target viewing geometries are some of these challenges. A common approach for target detection is to apply atmospheric correction algorithms to the radiance image data cube and then search within the atmospherically corrected image cube for the target reflectance signature of interest. One major issue with the above approach is that it is com-putationally demanding. In this paper, instead of applying atmospheric correction to the raw radiance data, we generate radiance profiles of burn scar for the observed atmospheric and illumination conditions at the time of the hyperspec-tral image data collection and form a radiance profile library using a nonlinear analytical model for radiative transfer and MODTRAN. The target detection has been performed by a spectral similarity technique which takes into consideration multiple radiance profile variants of the target of interest. The effectiveness of the radiance domain-based target detection approach on reducing the computation time has been demonstrated on burn scar detection using airborne AVIRIS image data.

NASA has been planning a hyperspectral infrared imager mission which will provide global coverage using a hyperspectral imager with 60-m resolution. In some practical applications, such as special crop monitoring or mineral mapping, 60-m resolution may still be too coarse. There have been many pansharpening algorithms for hyperspectral images by fusing high-resolution (HR) panchromatic or multispectral images with low-resolution (LR) hyperspectral images. We propose an approach to generating HR hyperspectral images by fusing high spatial resolution color images with low spatial resolution hyperspectral images. The idea is called hybrid color mapping (HCM) and involves a mapping between a high spatial resolution color image and a low spatial resolution hyperspectral image. Several variants of the color mapping idea, including global, local, and hybrid, are proposed and investigated. It was found that the local HCM yielded the best performance. Comparison of the local HCM with 10 state-ofthe-art algorithms using five performance metrics has been carried out using actual images from the air force and NASA. Although our HCM method does not require a point spread function (PSF), our results are comparable to or better than those methods that do require PSF. More importantly, our performance is better than most if not all methods that do not require PSF. After applying our HCM algorithm, not only the visual performance of the hyperspectral image has been significantly improved, but the target classification performance has also been improved. Another advantage of our technique is that it is very efficient and can be easily parallelized. Hence, our algorithm is very suitable for real-time applications.

The Reed–Xiaoli (RX) algorithm has been widely used as an anomaly detector for hyperspectral images. Recently, kernel RX (KRX) has been proven to yield high performance in anomaly detection and change detection. In this paper, we present a generalization of the KRX algorithm. The novel algorithm is called cluster KRX (CKRX), which becomes KRX under certain conditions. The key idea is to group background pixels into clusters and then apply a fast eigendecomposition algorithm to generate the anomaly detection index. Both global and local versions of CKRX have been implemented. Application to anomaly detection using actual hyperspectral images is included. In addition to anomaly detection , the CKRX algorithm has been integrated with other prediction algorithms for change detection. Spatially registered visible and near-infrared hyperspectral images collected from a tower-based geometry have been used in the anomaly and change detection studies. Receiver operating characteristics curves and actual computation times were used to compare different algorithms. It was demonstrated that CKRX has comparable detection performance as KRX, but with much lower computational requirements. Index Terms—Anomaly detection, change detection algorithms , cluster kernel RX (CKRX), hyperspectral imaging, kernel RX (KRX), receiver operating characteristics (ROC), Reed–Xiaoli (RX).

Hyperspectral images have been used in anomaly and change detection applications such as search and rescue operations where it is critical to have fast detection. However, conventional Reed-Xiaoli (RX) algorithm [6] took about 600 seconds using a PC to finish the processing of an 800x1024 hyperspectral image with 10 bands. This is not acceptable for real-time applications. A more recent algorithm known as kernel RX (KRX) [7] achieves better detection performance than RX at the expense of computational cost. For example, for the same 800x1024 image with 10 bands, KRX took 15 hours to finish the processing. In this paper, we present a general framework for fast anomaly detection using RX and KRX algorithms. First, a fast data reduction scheme using Principal Component Analysis (PCA) is proposed. This method takes less than 1 second to finish and the performance degradation is minimal. Second, we propose several speed boosting options in the RX and KRX algorithms. These options include image sub-sampling, the use of block pixels, and background pixel sub-sampling. Actual hyperspectral image has been used in our studies. Receiver operating characteristics (ROC) curves and actual computation times were used to compare the various options. For the 800x1024x10 image, we were able to improve the speed by more than 220 times for RX and 700 times for KRX with minimal degradation in detection performance.

Past estimates of chemical element concentration from investigations on the surface of Mars by Alpha-Particle X-ray Spectrometer (APXS) instrument have been conducted by converting the peak areas of the characteristic element lines into element concentrations using look-up calibration tables. In this work, we have investigated the feasibility of applying a linear spectral unmixing technique, NCLS (Nonnegatively Constrained Least Squares) to APXS spectra data for concentration estimation. A procedure for signature calibration with the NCLS technique is also introduced in this work. Estimates using NCLS are highly accurate in comparison to the applied benchmark technique, PLS (Partial Least Squares) in a leave-one-out testing framework that uses 11 geostandards. Future work will consider the performance comparison with a peak-area based concentration estimation method, and whether the fusion of the two methods can further increase the correct concentration estimation accuracy.

Con Edison experiences more than 1000 arcing faults on its secondary distribution system each year. Arcing faults introduce strong harmonics into the power network. We propose a fast, low-cost, and high-performance approach to locating arcing faults based on harmonics. First, we implemented two novel algorithms. One is based on the voltage ratio of harmonics. This method can detect arcing fault one at a time. The second one is based on sparse sensing, which is a powerful technique that can detect multiple faults and is robust to measurement noise. Both methods require low-cost voltage measurements; no high-cost sensors are required. In addition, the computations can be done very quickly within a few cycles (< 50 ms). Second, we have performed extensive simulation studies using the IEEE 14-bus system, IEEE 18-bus system, IEEE 118-bus system, and a 454-bus system. We only need to measure the voltages of a small percentage of the nodes. For example, voltages from only 20% of the nodes in a 454-bus system are needed for accurate fault location. Multiple simultaneous faults can be located. All the results clearly demonstrated the location accuracy of our algorithms.

Motor failures in aerospace applications can lead to serious compromises in safety, overall effectiveness, and maintenance costs. In mission critical applications, it is important that motor fault signatures are identified before a failure occurs. It is known that 40% of mechanical failures occur due to bearing faults. Bearing faults can be identified from the motor vibration signatures. Three key contributions are outlined in this paper. First, we develop a low cost test bed for simulating bearing faults in a motor. Second, we develop a wireless sensor module for collection of vibration data from the test bed. Finally, we use a novel two stage neural network to classify various bearing faults using the Generalized Hebbian Algorithm (GHA) in the first stage and a supervised learning vector quantization network (SLVQ) with a self organizing map approach for fault classification in the second stage.

Low-cost and portable gas chromatography-ion mobility spectrometry (GC-IMS) has been used to identify chemicals. To accomplish this, two parameters are used. The first parameter relates to the GC retention time (RT), which is the residence time of an analyte as it passes through the column. Different chemicals have different RTs. The second parameter is the drift time of ionized species derived for a specific chemical in the IMS. Due to molecular cross section, mass, and chemical properties, different chemicals produce ionized species with different drift times. Combining these two parameters, GC-IMS has been shown to distinguish between different chemicals. Chemical detection and identification are not that easy in practice. First, the concentration of chemicals may be very low, and it may be difficult to determine the chromatographic RT and IMS drift time for chemicals under these conditions. Second, the specific ionized species produced in the IMS are concentration dependent and the IMS spectra obtained at different analyte concentrations are not easily predictable. For example, at low concentrations, chemicals seldom form dimers following atmospheric pressure ionization. The possible presence of either monomers or dimers in the IMS drift tube may confuse the chemical classification process. Third, it is important to estimate the concentration of chemicals, as this information will provide toxicity, and the linear dynamic range of typical IMS systems is relatively low In this study, an image processing approach to enhancing the GC-IMS signal quality is introduced. The key idea in this approach is to treat GC-IMS data as an image and then apply an anomaly detector to detect and enhance abnormal regions in the image. The results of a study that compares a conventional approach to chemical detection and the introduced image enhancement approach are presented. Receiver operating characteristics curves were used to compare the detection performances of the two approaches.

Switchgear arcing faults have been a primary cause for concern for the manufacturing industry and safety personnel alike. The deregulation of the power industry being in full swing and the ever-growing competitiveness in the distribution sector call for the transition from preventive to predictive maintenance. Switchgears form an integral part of the distribution system in any power system setup. Keeping in mind the switchgear arcing faults, the aforementioned transition applies, most of all, to the switchgear industry. Apart from the fact that it is the primary cause of serious injuries to electrical workers worldwide, switchgear arcing faults directly affect the quality and continuity of electric power to the consumers. A great amount of technological advancement has taken place in the development of arc-resistant/proof switchgears. However, most of these applications focus on minimizing the damage after the occurrence of the arcing fault. The problem associated with the compromise on the quality and continuity of electric power in such a scenario still awaits a technical as well as economically feasible solution. This paper describes the development of a novel approach for the detection of arcing faults in medium-/low-voltage switchgears. The basic concept involves the application of differential protection for the detection of any arcing within the switchgear. The new approach differs from the traditional differential concept in the fact that it employs higher frequency harmonic components of the line current as the input for the differential scheme. Actual arc-generating test benches have been set up in the Power System Simulation Laboratory at the Energy Systems Research Center to represent both medium- and low-voltage levels. Hall effect sensors in conjunction with data acquisition in LabVIEW are employed to record the line current data before, during, and after the arcing phenomenon. The methodology is first put to test via simulation approach for medium-vol-\n-\ntage levels and then corroborated by actual hardware laboratory testing for low-voltage levels. The plots derived from the data gathering and simulation process clearly underline the efficiency of this approach to detect switchgear arcing faults. Both magnitude and phase differential concepts seem to provide satisfactory results. Apart from the technical efficiency, the approach is financially feasible, considering the fact that the differential protection is already being comprehensively employed worldwide.

An image texture analysis and target recognition approach of using an improved image texture feature coding method (TFCM) and Support Vector Machine (SVM) for target detection is presented. With our proposed target detection framework, targets of interest can be detected accurately. Cascade-Sliding-Window technique was also developed for automated target localization. Application to mammogram showed that over 88% of normal mammograms and 80% of abnormal mammograms can be correctly identified. The approach was also successfully applied to Synthetic Aperture Radar (SAR) and Ground Penetrating Radar (GPR) images for target detection.

Concentric ring arrays can provide effective beamforming and achieve frequency invariant beampatterns. For long range signal acquisition, the array has a large number of array elements, and partial adaptation is often necessary to increase tracking ability and reduce computation. The topic of this paper is the study of a partially adaptive concentric ring array for three-dimensional audio signal acquisition. We develop the partially adaptive array through partition matrix formulation, provide the associated adaptive structure, and derive the steady state residual interference and noise power that can serve as a criterion to evaluate different partition structures. A comparison of several partition schemes using the criterion is given, and the theoretical results are supported by simulations.

Electronic noses (e-noses) are commonly used to monitor air contaminants in space stations and shuttles. Data preprocessing (measurement denoising and feature extraction) and pattern classification are important problems of an e-nose system. In this paper, the application of a wavelet-based denoising method and a Dempster-Shafer(DS) classification fusion method in an e-nose system are proposed. Six transient-state features are extracted from the sensor measurements filtered by the wavelet denoising method and are used to train multiple classifiers such as multilayer perceptrons (MLPs), support vector machines (SVMs), k-nearest neighbors (KNNs), and the Parzen classifier. The DS technique is used at the end to fuse the results of the multiple classifiers to get the final classification. Experimental analysis based on real vapor data shows that the wavelet denoising method can successfully remove both random noise and outliers, and the classification rate can be improved by using classifier fusion.

This work focuses on an ultrasonic guided wave structural health monitoring (SHM) system development for aircraft wing inspection. In part I of the study, a detailed description of a real aluminum wing specimen and some preliminary wave propagation tests on the wing panel are presented. Unfortunately, strong attenuation and scattering impede guided waves for large-area inspection. Nevertheless, small, low-cost and lightweight piezoelectric (PZT) discs were bonded to various parts of the aircraft wing, in a form of relatively sparse arrays, for simulated cracks and corrosion monitoring. The PZT discs take turns generating and receiving ultrasonic guided waves. Pair-wise through-transmission waveforms collected at normal conditions served as baselines, and subsequent signals collected at defected conditions such as rivet cracks or corrosion detected the presence of a defect and its location with a novel correlation analysis based technique called RAPID (reconstruction algorithm for probabilistic inspection of defects). The effectiveness of the algorithm was tested with several case studies in a laboratory environment. It showed good performance for defect detection, size estimation and localization in complex aircraft wing structures. (Some figures in this article are in colour only in the electronic version)

The objective of this study is to develop a wireless ultrasonic structural health monitoring (SHM) system for aircraft wing inspection. In part I of the study (Zhao et al 2007 Smart Mater. Struct. 16 1208–17), small, low cost and light weight piezoelectric (PZT) disc transducers were bonded to various parts of an aircraft wing for detection, localization and growth monitoring of defects. In this part, two approaches for wirelessly interrogating the sensor/actuator network were developed and tested. The first one utilizes a pair of reactive coupling monopoles to deliver 350 kHz RF tone-burst interrogation pulses directly to the PZT transducers for generating ultrasonic guided waves and to receive the response signals from the PZTs. It couples enough energy to and from the PZT transducers for the wing panel inspection, but the signal is quite noisy and the monopoles need to be in close proximity to each other for efficient coupling. In the second approach, a small local diagnostic device was developed that can be embedded into the wing and transmit the digital signals FM-modulated on a 915 MHz carrier. The device has an ultrasonic pulser that can generate 350 kHz, 70 V tone-burst signals, a multiplexed A/D board with a programmable gain amplifier for multi-channel data acquisition, a microprocessor for circuit control and data processing, and a wireless module for data transmission. Power to the electronics is delivered wirelessly at X-band with an antenna–rectifier (rectenna) array conformed to the aircraft body, eliminating the need for batteries and their replacement. It can effectively deliver at least 100 mW of DC power continuously from a transmitter at a range of 1 m. The wireless system was tested with the PZT sensor array on the wing panel and compared well with the wire connection case. (Some figures in this article are in colour only in the electronic version)

— Automated operations based on voice commands will become more and more important in many applications, including robotics, maintenance operations, etc. However, voice command recognition rates drop quite a lot under non-stationary and chaotic noise environments. In this paper, we tried to significantly improve the speech recognition rates under non-stationary noise environments. First, 298 Navy acronyms have been selected for automatic speech recognition. Data sets were collected under 4 types of noisy environments: factory, buccaneer jet, babble noise in a canteen, and destroyer. Within each noisy environment, 4 levels (5 dB, 15 dB, 25 dB, and clean) of Signal-to-Noise Ratio (SNR) were introduced to corrupt the speech. Second, a new algorithm to estimate speech or no speech regions has been developed, implemented, and evaluated. Third, extensive simulations were carried out. It was found that the combination of the new algorithm, the proper selection of language model and a customized training of the speech recognizer based on clean speech yielded very high recognition rates, which are between 80% and 90% for the four different noisy conditions. Fourth, extensive comparative studies have also been carried out.

The application of active flow control via synthetic jet actuators for separation and roll control on a scaled Cessna 182 model was investigated experimentally in a low-speed wind tunnel. The model was instrumented with either ailerons or synthetic jets embedded within the outer portion of the wings' span, in lieu of ailerons. Force and moment measurements were performed for various aileron deflections and synthetic jet momentum coefficients (on either both wingtips or only on one). It was found that the effectiveness of the synthetic jets is comparable to that of conventional ailerons at moderate deflection angles. The model was also instrumented with a hot-film shear stress sensor downstream from the synthetic jet exit. The sensor's rms output was monitored in real time by a computer. When the rms reached a predetermined threshold value, the computer automatically turned on the synthetic jet actuators. Using the appropriate threshold value resulted in complete avoidance of wingtip stall at the angle of attack where separation would have occurred. In addition, the shear stress sensor and wind tunnel force data were used to identify the system dynamics. A computerized dynamic model of an RC version of the Cessna 182 showed that at moderate to high angles of attack, synthetic jets alone could be used to control the roll of the aircraft. Nomenclature b = wing span C D = vehicle drag coefficient C L = vehicle lift coefficient C m = vehicle pitching moment (about quarter-chord) C m = derivative of C m with respect to the angle of attack C n = vehicle yaw moment coefficient (about quarter-chord and center plane) C r = vehicle rolling moment coefficient C = synthetic jets' momentum coefficient, …n j j U 2 j hL j †=… 1 2 1 U 2 1 cb† c = mean chord f r …V†, f n …V† = coefficients listed in Eqs. (4) and (5) g = acceleration due to gravity h = synthetic jet orifice width L j = synthetic jet orifice length m = mass n j = number of active synthetic jets Re = Reynolds number, …U 1 c†= T = reattachment time constant T f = time of flight, c=U 1 T j = synthetic jet period t = time U j = synthetic jet orifice velocity (during the outstroke), 1 T j R 0 u j …t† dt U 1 = freestream velocity u b = x-component of inertial velocity in airplane body frame F B u j …t† = phase-averaged velocity at the jet exit plane V = input voltage to the synthetic jets (before amplification) w b = z-component of inertial velocity in airplane body frame F B x sj = streamwise location of the synthetic jet actuators = angle of attack C r = change in rolling moment coefficient with respect to aileron deflection of 3 deg C r sj = change in rolling moment coefficient from synthetic jet configuration (jets off) a = aileron deflection E = elevator command R = rudder command SV = synthetic jet command Th = engine command = closed-loop control system characteristics root j = synthetic jet density 1 = freestream fluid density = synthetic jet blowing time, T j =2

—The current speech interfaces in many military applications may be adequate for native speakers. However, the recognition rate drops quite a lot for non-native speakers (people with foreign accents). This is mainly because the non-native speakers have large temporal and intra-phoneme variations when they pronounce the same words. This problem is also complicated by the presence of large environmental noise such as tank noise, helicopter noise, etc. In this paper, we proposed a novel continuous acoustic feature adaptation algorithm for on-line accent and environmental adaptation. Implemented by incremental singular value decomposition (SVD), the algorithm captures local acoustic variation and runs in real-time. This feature-based adaptation method is then integrated with conventional model-based maximum likelihood linear regression (MLLR) algorithm. Extensive experiments have been performed on the NATO non-native speech corpus with baseline acoustic model trained on native American English. The proposed feature-based adaptation algorithm improved the average recognition accuracy by 15%, while the MLLR model based adaptation achieved 11% improvement. The corresponding word error rate (WER) reduction was 25.8% and 2.73%, as compared to that without adaptation. The combined adaptation achieved overall recognition accuracy improvement of 29.5%, and WER reduction of 31.8%, as compared to that without adaptation.

This work focuses on an ultrasonic guided wave structural health monitoring (SHM) system development for aircraft wing inspection. In part I of the study, a detailed description of a real aluminum wing specimen and some preliminary wave propagation tests on the wing panel are presented. Unfortunately, strong attenuation and scattering impede guided waves for large-area inspection. Nevertheless, small, low-cost and light-weight piezoelectric (PZT) discs were bonded to various parts of the aircraft wing, in a form of relatively sparse arrays, for simulated cracks and corrosion monitoring. The PZT discs take turns generating and receiving ultrasonic guided waves. Pair-wise through-transmission waveforms collected at normal conditions served as baselines, and subsequent signals collected at defected conditions such as rivet cracks or corrosion detected the presence of a defect and its location with a novel correlation analysis based technique called RAPID (reconstruction algorithm for probabilistic inspection of defects). The effectiveness of the algorithm was tested with several case studies in a laboratory environment. It showed good performance for defect detection, size estimation and localization in complex aircraft wing structures.

The application of active flow control via synthetic jet actuators for separation and roll control on a scaled Cessna 182 model was investigated experimentally in a low-speed wind tunnel. The model was instrumented with either ailerons or synthetic jets embedded within the outer portion of the wings’span, in lieu of ailerons. Force and moment measurements were performed for various aileron deflections and synthetic jet momentum coefficients (on either both wingtips or only on one). It was found that the effectiveness of the synthetic jets is comparable to that of conventional ailerons at moderate deflection angles. The model was also instrumented with a hot-film shear stress sensor downstream from the synthetic jet exit. The sensor’s rms output was monitored in real time by a computer. When the rms reached a predetermined threshold value, the computer automatically turned on the synthetic jet actuators. Using the appropriate threshold value resulted in complete avoidance of wingtip stall at the angle of attack where separation would have occurred. In addition, the shear stress sensor and wind tunnel force data were used to identify the system dynamics. A computerized dynamic model of an RC version of the Cessna 182 showed that at moderate to high angles of attack, synthetic jets alone could be used to control the roll of the aircraft.

A complete system was built for high-performance image compression based on overlapped block transform. Extensive simulations and comparative studies were carried out for still image compression including benchmark images (Lena and Barbara), synthetic aperture radar (SAR) images, and color images. We have achieved consistently better results than three commercial products in the market (a Summus wavelet codec, a baseline JPEG codec, and a JPEG-2000 codec) for most images that we used in this study. Included in the system are two post-processing techniques based on morphological and median filters for enhancing the perceptual quality of the reconstructed images. The proposed system also supports the enhancement of a small region of interest within an image, which is of interest in various applications such as target recognition and medical diagnosis.

—This correspondence proposes a method for array pattern synthesis of a concentric ring array that yields invariant array pattern over a certain frequency band for beamforming in three dimensions. The proposed method uses the Fourier–Bessel series to determine the weights from different rings to form the overall array pattern. Simulation examples are given to illustrate the proposed method.

—This correspondence proposes a method for array pattern synthesis of a concentric ring array that yields invariant array pattern over a certain frequency band for beamforming in three dimensions. The proposed method uses the Fourier–Bessel series to determine the weights from different rings to form the overall array pattern. Simulation examples are given to illustrate the proposed method.

Real-time speaker verification, with speech acquired using the NIST Mk-III microphone array and an autodirective beamforming algorithm, is demonstrated. The software and hardware backbone of the demonstration is the NIST Smart Flow System and Mk-III Array, both developed by National Institute of Standards and Technology in support of multimodal research communities. A microphone array acquires speech signals; a steered response

A complete system was built for high-performance image compression based on overlapped block transform. Extensive simulations and comparative studies were carried out for still image compression including benchmark images (Lena and Barbara), synthetic aperture radar (SAR) images, and color images. We have achieved consistently better results than three commercial products in the market (a Summus wavelet codec, a baseline JPEG codec, and a JPEG-2000 codec) for most images that we used in this study. Included in the system are two post-processing techniques based on morphological and median filters for enhancing the perceptual quality of the reconstructed images. The proposed system also supports the enhancement of a small region of interest within an image, which is of interest in various applications such as target recognition and medical diagnosis.

This correspondence proposes a method for array pattern syn-thesis of a concentric ring array that yields invariant array pattern over a certain frequency band for beamforming in three dimensions. The pro-posed method uses the Fourier–Bessel series to determine the weights from different rings to form the overall array pattern. Simulation examples are given to illustrate the proposed method.

In this paper, spectral unmixing methods, which are extensively used in hyperspectral imaging area, are proposed for classification and abundance fraction (concentration) estimation of chemical and biological agents that exist in the mixture form. Several government-furnished datasets, which were collected through the infrared spectrum method, were thoroughly analyzed. Two similarity measures-the spectral angle mapper and spectral information divergence-were investigated in order to provide a quantitative comparison basis with respect to the performance of the applied spectral unmixing methods in the existence of similar and distinct agents. The use of the similarity measures provided valuable information about the signature characteristics of the agents, which led to a better understanding about the capabilities of the investigated methods. The orthogonal subspace projection (OSP) method was investigated as the first unmixing, classification, and abundance estimation technique. It was observed that the OSP method provided good results when the number of agents in the database was small and was composed of distinct agents. However, when the number of agents was incremented by adding agents that share similar characteristics, the abundance estimation accuracy gradually degraded in addition to generating negative abundance fraction estimates. The second investigated unmixing method was called nonnegatively constrained least squares (NCLS). The results and analyses indicated that the NCLS method outperformed the OSP approach by providing considerably more accurate fraction estimates while at the same time not generating any negative fraction estimates; thus, the use of the NCLS method was found to be promising in detection and abundance fraction estimation of chemical and biological agents that exist in the form of mixtures. In addition, efficient implementation of NCLS has resulted in much lower computations than the conventional OSP implementation.

avoid bird strikes to aircraft. First, a cost-effective microphone dish concept (microphone array with many concentric rings) is presented that can provide directional and accurate acquisition of bird sounds and can simultaneously pick up bird sounds from different directions. Second, direction-of-arrival (DOA) and beamforming algorithms have been developed for the circular array. Third, an efficient recognition algorithm is proposed which uses Gaussian mixture models (GMMs). The overall system is suitable for monitoring and recognition for a large number of birds. Fourth, a hardware prototype has been built and initial experiments demonstrated that the array can acquire and classify birds accurately.

Circumferential guided ultrasonic Shear Horizontal (SH) wave Electromagnetic Acoustic Transducer (EMAT) pairs mounted on a mobile fixture in a through-transmission mode were used for detection and characterization of mechanical dents on the outer surface of a pipe wall from inside the pipe. The dents were created on a 12 in. diameter standard seamless steel pipe by hydraulically pressing steel balls of various sizes into the pipe wall. n 1 mode SH wave was directed through and along the wall of the pipe. Multiple measurements were obtained both from the dents and from the no-flaw region of the pipe using the EMAT pair. Dent features were extracted with a Principal Component Analysis (PCA) technique and classified into " cup " and " saucer " types using Discriminant Analysis (DA). The overall approach is able to detect and classify dents of depth 25% through wall or deeper, which should meet the needs of the pipeline safety inspection community

Circumferential guided ultrasonic Shear Horizontal (SH) wave Electromagnetic Acoustic Transducer (EMAT) pairs mounted on a mobile fixture in a through-transmission mode were used for detection and characterization of mechanical dents on the outer surface of a pipe wall from inside the pipe. The dents were created on a 12-inch diameter standard seamless pipe by hydraulically pressing steel balls of various sizes into the pipe wall. n1 mode SH wave was directed through and along the wall of the pipe. Multiple measurements were obtained from the dents and no-flaw region of the pipe using the EMAT pair. Dent features were extracted with a Principal Component Analysis (PCA) technique and classified into "cup" and "saucer" types using Discriminant Analysis (DA). The overall approach is able to detect and classify dents of depth 2.5mm (0.1 inch) or larger, which should meet the needs of the pipeline safety inspection community [1]. Preliminary dent depth estimation potential is also shown via an amplitude correlation approach.

This paper documents the experimental validation of an active control approach for mitigating chatter in milling. To the authors knowledge, this is the first successful hardware demonstration of this approach. This approach is very different from many existing approaches that avoid instabilities by varying process parameters to seek regions of stability or by altering the regenerative process. In this approach, the stability lobes of the machine and tool are actively raised. This allows for machining at spindle speeds that are more representative of those used in existing machine tools. An active control system was implemented using actuators and sensors surrounding a spindle and tool. Due to the complexity of controlling from a stationary coordinate system and sensing from a rotating system, a telemetry system was used to transfer structural vibration data from the tool to a control processor. Closed-loop experiments produced up to an order of magnitude improvement in metal removal rate.

—Sensor and actuator self-validation is a critical step in system control and fault diagnostics. If sensors do not work properly, one cannot rely on their outputs to further deduce system status. Similarly, faulty actuators will not satisfy system performance objectives and may cause disasters in feedback control systems. In this paper, a novel method to generate structured residuals for isolating sensor and actuator failures with the least sensitivity to model-plant-mismatch (MPM) and disturbances in multivariate dynamic systems is proposed. The proposed method includes two components. The first component is the generation of the primary residuals directly from noisy input and output measurements without identifying explicitly the model of a system under consideration. The primary residuals are generated such that they have the least sensitivity to any MPM and process disturbances , but have the highest sensitivity to faults in any sensors and/or actuators. The second component of the proposed scheme is the max–min design to transform the primary residuals into a set of structured residuals for fault isolation by improving the existing structured residual approach with maximized sensitivity (SRAMS) [18]. Since one structured residual is made immune to a specified subset of faults, but very sensitive to other faults, any faulty sensors and/or actuators can be isolated by observing the structured residuals in accordance with a predetermined isolation logic. The proposed method has been verified for detection and isolation of faulty sensors and/or actuators in an experimental pilot plant.

This paper documents the experimental validation of an active control approach for mitigating chatter in milling. To the authors knowledge, this is the first successful hardware demonstration of this approach. This approach is very different from many existing approaches that avoid instabilities by varying process parameters to seek regions of stability or by altering the regenerative process. In this approach, the stability lobes of the machine and tool are actively raised. This allows for machining at spindle speeds that are more representative of those used in existing machine tools.An active control system was implemented using actuators and sensors surrounding a spindle and tool. Due to the complexity of controlling from a stationary co-ordinate system and sensing from a rotating system, a telemetry system was used to transfer structural vibration data from the tool to a control processor. Closed-loop experiments produced up to an order of magnitude improvement in metal removal rate.

In this paper, we summarise our recent results in fault tolerant formation control of Unmanned Air Vehicles (UAV). A fault tolerant control scheme to deal with both GPS sensor failure and wireless communication packet losses is presented. Moreover, some extensions to the formation control algorithms developed by UC Berkeley are made to support time-varying heading and curved flight trajectories. The effectiveness of the presented fault tolerant control scheme is illustrated by real-time formation flight simulations conducted in a wireless network environment. Both rotary and fixed wing UAVs, mesh and triangular formations, and straight and curved trajectories are considered in our real-time simulations.

A prototype wireless guided wave inspection system is realized by using a station monopole antenna as a transmitter, an on-board antenna as transponder, a PVDF comb transducer for generating and receiving ultrasonic Lamb waves in a layered structure, and another portable active monopole antenna as a receiver. Experiments on a 0.8mm thick aluminum plate with a 12mm long, 50% through-the-wall crack clearly showed its feasibility on defect detection. The conventional wired approach and the interim semi-wireless approach are also presented for comparison purpose. Practical leave-on-board antenna designs for aircraft wing are also discussed, along with some preliminary experiments.

In this paper, a novel neural network (NN) backstep-ping controller is modified for application to an industrial motor drive system. A control system structure and NN tuning algorithms are presented that are shown to guarantee stability and performance of the closed-loop system. The NN backstepping controller is implemented on an actual motor drive system using a two-PC control system developed at The University of Texas at Arlington. The implementation results show that the NN backstepping controller is highly effective in controlling the industrial motor drive system. It is also shown that the NN controller gives better results on actual systems than a standard backstepping controller developed assuming full knowledge of the dynamics. Moreover, the NN controller does not require the linear-in-the-parameters assumption or the computation of regression matrices required by standard backstepping.

Sensor self-validity check is a critical step in system control and fault diagnostics. In this paper, a robust approach to isolate sensor failures is proposed. First, a residual model for a given system is built off-line and directly based on input-output measurement data. The residual model outputs are called " primary residuals " and are zero when there is no fault. Most conventional approaches to residual model generation are indirect, as they first require the determination of state-space or other models using standard system identification algorithms. Second, a new max-min design of structured re-siduals, which can maximize the sensitivity of structured residuals with respect to sensor failures, is proposed. Based on the structured residuals, one can then isolate the sensor failures. This design can also be done in an off-line manner. It is an optimization procedure that avoids local optimal solutions. Simulation and experimental results demonstrated the effectiveness of the proposed method.

A desired compensation adaptive law-based neural network (DCAL-NN) controller is proposed for the robust position control of rigid-link robots. The NN is used to approximate a highly nonlinear function. The controller can guarantee the global asymptotic stability of tracking errors and boundedness of NN weights. In addition, the NN weights here are tuned on-line, with no off-line learning phase required. When compared with standard adaptive robot controllers, we do not require linearity in the parameters, or lengthy and tedious preliminary analysis to determine a regression matrix. The controller can be regarded as a universal reusable controller because the same controller can be applied to any type of rigid robots without any modifications. A comparative simulation study with different robust and adaptive controllers is included.

This note presents an approach to exploit symmetry in the synthesis of solutions for the SCP. The basic idea underlying the method is to perform the computation of the maximal controllable sublanguage of a given language over reduced automaton representations of the plant and specification. The reduction in state spaces can be significant when the degree of symmetry is high. In general the determination of symmetric groups of languages is a very complex task. When dealing with systems with replicated structures, however, the symmetric group emerges naturally as being formed by the rotations over the similar components. Besides the gain in computational complexity, our approach also provides a way to implement supervision using the reduced automaton representation for the supervisor instead of its corresponding expanded representation. We are currently investigating methods for relaxing the notion of symmetry considered here, in order to use the idea of reduced automata for a larger class of problems. Some preliminary results of this work can be found in [6]. Also, motivated by the robustness problem studied in [2], we are investigating the possibility of using the idea of quotient structure to derive results where we could implement a robust parametrized supervisor as the solution for a class of different plants. Abstract—Zak and Hui [1] proposed a sliding mode controller for linear multiple-input–multiple-output (MIMO) systems using static output feedback. The note in [2] provided an improvement of the output feedback controller in [1] for a class of linear single-input–single-output (SISO) systems that eliminated two important limitations of [1]: (a) system uncertainties must be bounded by the system output; and (b) a requirement of a matrix inequality [1, eq. (4.3)]. The controller in [2] can guarantee global closed-loop stability. This note extends the results of [2] to linear MIMO systems. It is emphasized that the proposed MIMO controller yields global closed-loop stability whereas the one in [1] can only guarantee local stability. An application of the proposed MIMO controller to an aircraft model is included to show the effectiveness of the method.

In the above paper, 1 some typographical errors appeared that need to be corrected. In (17), the sign = should be dropped. The (2; 2) block of matrix 51 at the bottom of page 494 and the matrix in (28) at the bottom of p. 495 should be 0m(1 + 2) 01 P: The first block of matrices B2 and B3 should be m0 and m, respectively. V (e; t) = ke(t)k 2 P + 0 0 t t+ 01 1 kNe(s)k 2 P ds + t t+0 01 2 kN d e(s)k 2 P ds d where ke(t)k 2 P = e T (t)Pe(t) and inequality (27) should be _ V (e; t) e T (t) (N + N d) T P + P (N + N d) + m (1 + 2)PN d P 01 N T d P + m (01 1)N T PN + m 01 2 N T d PN d e(t)

—A controller is proposed for the robust backstepping control of a class of general nonlinear systems using neural networks (NNs). A new tuning scheme is proposed which can guarantee the boundedness of tracking error and weight updates. Compared with adaptive backstepping control schemes, we do not require the unknown parameters to be linear parametrizable. No regression matrices are needed, so no preliminary dynamical analysis is needed. One salient feature of our NN approach is that there is no need for the off-line learning phase. Three nonlinear systems, including a one-link robot, an induction motor, and a rigid-link flexible joint robot, were used to demonstrate the effectiveness of the proposed scheme.

—In this paper, we present a new robust control technique for induction motors using neural networks (NNs). The method is systematic and robust to parameter variations. Motivated by the well-known backstepping design technique, we first treat certain signals in the system as fictitious control inputs to a simpler subsystem. A two-layer NN is used in this stage to design the fictitious controller. Then we apply a second two-layer NN to robustly realize the fictitious NN signals designed in the previous step. A new tuning scheme is proposed which can guarantee the boundedness of tracking error and weight updates. A main advantage of our method is that we do not require regression matrices, so that no preliminary dynamical analysis is needed. Another salient feature of our NN approach is that the off-line learning phase is not needed. Full state feedback is needed for implementation. Load torque and rotor resistance can be unknown but bounded.

In this paper, we propose to use adaptive fuzzy logic to tackle the spacecraf t attitude control problem. T he advantage is that no linearity in the system param eter assumption is needed. An on-line tuning schem e with no oo-line training phase is used to update the weights in the fuzzy logic controller. Attitude tracking errors are guaranteed to be bounded.

A robust neural network NN controller is proposed for the simultaneous forcermo-tion control of constrained rigid robots. The NN weights here are tuned on-line, with no off-line learning phase required. Most importantly, we can guarantee the boundedness of constraint force errors, joint position tracking errors, and NN weights. When compared with adaptive controllers, we do not require linearity in the unknown parameters, and the tedious computation of the regression matrix. Novel passivity properties of the NN controller are stated and proven.

A robust Neural Network (NN) controller is proposed for the motion control of rigid-link flexible-joint (RLFJ) robots. No weak joint elasticity assumption is needed. The NNs are used to approximate three very complicated nonlinear functions. Our NN approach requires no off-line learning phase, and no lengthy and tedious preliminary analysis to find the regression matrices. Most importantly, we can guarantee the uniformly ultimately bounded (UUB) stability of tracking errors and NN weights. The controller can be regarded as a universal reusable controller because the same controller can be directly applied to different RLFJ robots with different masses and lengths within the same class, for instance, of two-link revolute RLFJ robots.

—The purpose of this paper is to point out a confusing phenomenon in the teaching of Kalman filtering. Students are often confused by noting that the a posteriori error covariance of the discrete Kalman filter (DKF) is smaller than the error covariance of the continuous Kalman filter (CKF), which would mean that the DKF is better than CKF since it gives a smaller error covariance. However, simulation results show that CKF gives estimates much closer to the true states. We will provide a simple qualitative argument to explain this phenomenon.

This paper applies the FR-operator technique to the robust
stability problem of sampled-data systems against additive/multiplicative
perturbations, where a reasonable class of perturbations consists of
unstable as well as stable ones. Assuming that the number of unstable
modes of the plant does not change, we show that a small-gain condition
in terms of the FR-operator representation (which is actually equivalent
to a small-gain condition in terms of the L2-induced norm) is still
necessary and sufficient for the sampled-data system to be robustly stable
against h-periodic perturbations, in spite of their possible instability. The
result is derived by a Nyquist-type of arguments. Next, a necessary and
sufficient condition for robust stability against linear time-invariant (LTI)
perturbations is also given. Furthermore, we show that if the plant is
either single-input or single-output, the condition can be reduced to a
readily testable form. Finally, we clarify when the small-gain condition
becomes a particularly poor measure for robust stability.

— A robust neural-network (NN) controller is proposed for the motion control of rigid-link electrically driven (RLED) robots. Two-layer NN's are used to approximate two very complicated nonlinear functions. The main advantage of our approach is that the NN weights are tuned on-line, with no off-line learning phase required. Most importantly, we can guarantee the uniformly ultimately bounded (UUB) stability of tracking errors and NN weights. When compared with standard adaptive robot controllers, we do not require lengthy and tedious preliminary analysis to determine a regression matrix. The controller can be regarded as a universal reusable controller because the same controller can be applied to any type of RLED robots without any modifications.

A new approach to traffic incident detection is proposed in this paper. The method consists of two stages. First, a real-time adaptive on-line procedure is used to extract the significant components of traffic states, namely, average velocity and density of moving vehicles. Second, we apply a new neural network called Fuzzy CMAC (Cerebellar Arithmetic Computer) to identify traffic incidents. Fuzzy CMAC is an ideal candidate for this purpose for the following reasons. First, the Fuzzy CMAC learning structure is a creative use of fuzzy logic and CMAC based neural networks. Expert knowledge in terms of linguistic rules can be incorporated into the design. Second, the learning process is well suited for real-time application since the training process is an order of magnitude faster than conventional neural nets. Third, the Fuzzy CMAC can be implemented in high speed, highly parallel hardware. The importance of this research is threefold. One is that a good traffic incident detection system will help drivers to select an optimum route. The second one is that the system will be able to provide information for efficient dispatching of emergency services. Lastly, it will provide accurate knowledge of existing traffic conditions in order to guide effective on-line traffic controls.

Consider the response of an unstable LTI filter to an input over a time interval [0; T ]. The total response consists of the signal response and the natural response of the filter. The latter grows exponentially so that an unstable filter is not usable. This note shows that by holding the final value of the filter fixed to a finite value (say, zero), the natural response can be forced to be negligible over some initial part of the time interval [0; T ] so that the filter can be used. A means of implementing this idea is to reverse the input signal across the interval [0; T ] as suggested in [1]–[3]. The purpose of this brief is to present a state–variable perspective of this technique. An example is shown to implement a filter with no phase shifts.

In this paper, we propose to use a new neural network called Fuzzy CMAC (Cerebellar Model Arithmetic Computer) to tackle the attitude control problem. The advantages are no linearity in the system parameter assumption is needed and no upper bounds of unknown parameters is required. An on-line tuning scheme with no off-line training phase is used to update the weights in the neural network. Attitude tracking errors are guaranteed to be bounded.

We proposed a new sliding control method using output feedback
that can guarantee global stability. The current scheme is very simple
in structure because no observer is needed. The computational
requirement is also very small. It is robust to all initial conditions,
mismatched disturbance and parametric uncertainties.

Three practical techniques-Fuzzy Logic (F' L), Neural Networks (NN), and Auto-regressive model (AR)-for very short-term load forecasting have been propwed and discussed in this paper. Their performances are evaluated through a simulation study. The preliminary study shows that it is feasible to design a simple, satisfactory dynamic forecaster to predict the very short-term load trends on-line. FL and NN can be good candidates for this application. Kevwords: AGC, load forecast, artificial intelligent

A new robust adaptive control scheme is proposed for the simultaneous force/motion control of constrained rigid robots including motor dynamics. When slow motor dynamics are present, two problems arise. First, the performance of the controller will be degraded because of the interaction between robot and motor. Second, the uncertainties in the robot are no longer in the range space of control, which means that conventional methods cannot be easily applied. To confront these problems, a novel sliding mode technique is proposed which can achieve robustness to parameter variations in both manipulator and motor. Also the joint position errors can be driven to zero and the force errors can be reduced to arbitrarily small values. No joint acceleration measurement is needed.

We present a new adaptive control technique for induction motors. We fhst treat certain signals in the system as fictitious control signals to a simpler subsystem. An adaptive technique is used in this stage to design the fictitious controllers. Then we apply the sliding mode method to robustly realiie the fictitious adaptive signals. The method is robust to parameter variations, and the stability analysis is simple. Another advantage of our method is that we only require a reduced-order system to be linearly parametrizable (LP); the rest of the system dynamics can be highly nonlinear, with no LP requirement. Full state feedback is not needed for implementation. In particular, the rotor flux measurement is not needed. Load torque and rotor resistance can be unknown but bounded.

In a recent work by Zak and Hui [1], a sliding-mode controller for multi-input multi-output (MIMO) systems using static output feedback was proposed. Very nice geometric conditions for how to design sliding surfaces were given. However, there are two restrictive assumptions in it. One is that the uncertainties in the system must be
bounded by a known function of outputs which excludes some possible uncertainties in the A matrix if the system is described by the triple (A, B, C) . The other one requires a matrix equality [1,( 4.3)] to be held which may also be very difficult to satisfy in many systems. In this paper, we propose a modification of the sliding mode controller for a class of single-input/single-output (SISO) systems which can eliminate the above mentioned limitations and, under certain conditions, guarantee global closed-loop stability. Hence the range of applicability of the method in [1] can be greatly broadened.

A robust neural network (NN) scheme is proposed for the coordination control of robots carrying the same object. The NN weights here are tuned on-line, with no off-line learning phase required. Most importantly, we can guarantee the boundness of internal force errors, position tracking errors, and NN weights. Also no exact knowledge of the robot dynamics is required so that the NN controller is applicable to any type of rigid robots. When compared with adaptive controllers, we do not require persistent excitation condition, linearity in the unknown system parameters, and the tedious computation of the regression matrix.

In the active chatter control of machine tools, the most effective way to suppress the chatter is to place the actuator as close as possible to the tool tip. However, in practice, it is almost impossible to put the actuator at the same location of the tool tip. Also, in many machines the cutting tools are usually long and may be flexible. Both of these problems pose serious problems in machine chatter control. In order to control the chatter effectively and efficiently, a systematic methodology is proposed in this paper to deal with the modeling and control system design aspects of this challenging problem. Because of the flexibility effect in the tool shaft, conventional active control approaches may not perform in an efficient and effective manner. Here, two advanced control algorithms (LMS adaptive filter and fuzzy CMAC neural network) are proposed to counteract this problem. Experimental results on a lathe machine are also included. Approximately 20 dB reduction in chatter has been achieved.

We would like to thank Dr Phadke for his interests in our brief paper. He raised three concerns. We should like to respond to these points in turn. Stability Dr Phadke suggested a different Lyapunov function (equation (3) of Phadke's (1995) comments) to analyze the system stability. We are aware of this result. The reason we did not use it is because the resulting control law can only guarantee o+O asymptotically. In other words, no sliding mode exists in finite time. This means that the control law is no longer a sliding control law. It is only a robust control law. The most important concern is about the stability of the system in the reaching phase of the sliding mode. Using our control law, we can guarantee that (T = 0 in finite time-i' > 0. Is it possible for the state xi and parameter estimate 0 to go to infinity in finite time T? We shall argue that this is impossible. We prove this by contradiction. Suppose either x, or 8 goes to infinity in finite time T. Then the Lyapunov function V goes to infinity. Integrating equation (2) of Phadek (1995) yields bTvdl=[: [-:x:Qx, + aA:&,] dr (1) remembering that cr is bounded owing to the use of our sliding controller. Hence the right-hand side of (1) is negurive, since the quadratic term dominates the linear term. However, the left-hand side is V(T)-V(0) = x (positive), which results in a contradiction. Therefore both xi and 8 are bounded. As a matter of fact, it is the usual practice in the sliding mode control literature to ignore the stability analysis of the reaching phase. Investigators concentrate only on the stability analysis of the closed-loop system in sliding mode. The above argument can be treated as a justification for this phenomenon. Chang and Hurmuzlu (1993) have suggested a Correspondence Item KWAN method to eliminate the reaching phase of sliding mode. We could adopt that approach to improve our paper. Adaptation law Dr Phadke correctly pointed out that we assumed 6 = 0 in our paper. In fact, we did use constant parameters in our simulation examples when we first submitted the paper. During the first refereeing process, one referee raised concern about the effectiveness of our method with respect to slowly varying parameters. That is why we inserted some time-varying parameters in the examples in the final version of the paper to show the effectiveness of our method. We did not explain this in our final paper. We thank Dr Phadke for bringing this up.

This note points out some errors in the above paper.' It is argued that the proof in the main theorem is not mathematically rigorous. The control scheme requires the derivative of the sliding variable to be available, which is equivalent to saying that the uncertainties of the system are measurable. This is highly impractical if not impossible to implement. The theorem also says that the proposed new switching surface and control scheme can guarantee robustness of the system from initial time to final time. This is also incorrect because the theorem only shows that the sliding variable tends to zero approximately, i.e., the sliding motion only occurs approximately at t=m. The paper also contains several misprints.

It is well-known that sliding mode control is robust to matched uncertainties that lie in the range space of the input matrix. However many systems are affected by mismatched uncertainties and yet do not enjoy the matching conditions. It is the purpose of this paper to present a new dynamical approach of sliding variable formulation which, when the system is in sliding mode, can explicidy deal with mismatched uncertainties. Our first step is to treat certain states as inputs to a reduced-order system and use adaptive techniques to design fictitious controllers for these inputs, which can then tackle the mismatched uncertainties. The second step is to use sliding control to realize the adaptive fictitious controllers. The design is systematic, modular and intuitively simple. 1. Introducdon Consider the regulation of the following controllable single-input system: *I = (All + ~~I)XI + (Arz + u,z)xz (la) xz = (A*, + AA2,)x, + (AZ2 + A&,)x, + (b + Ab)u, b > 0, (lb) where u, xz, AZ2 are in R' x,, A,,, A$ are in R"-' A,, is in R(n-')x(n-').

Abstmct-Based on a decompasition of the rigid robot system with motor dynamics, a novel sliding-adaptive controller is developed which ean achieve robustness to parameter variations in both manipulator and motor. When the system is in sliding mode, force, position, and redundant joint velodty e m r s will approach zero irrespective of parametric uncertainties. Unlike conventional sliding techniques which are only robust to matched uncertainties, the proposed sliding method is robust to both matched and "ptehed uncertainties. Hence the scope of applications of sliding mode method can be broadened to nonlinear systems with mismatched parameter variations. No joint acceleration measurement is needed.

An open-loop state space model of all the major low-level rf feedback control loops is derived. The model has control and state variables for fast-cycling machines to apply modem multivariable feedback techniques. A condition is derived to know when exactly we can cross the boundaries between time-varying and time-invariant approaches for a fast-cycling machine like the Low Energy Booster (LEB). The conditions are dependent on the Q of the cavity and the rate at which the frequency changes with time. Apart from capturing the time-variant characteristics, the errors in the magnetic field are accounted in the model to study the effects on synchronization with the Medium Energy Booster (MEB). The control model is useful to study the effects on beam control due to heavy beam loading at high intensities, voltage transients just after injection especially due to time-varying voltages, instability thresholds created by the cavity tuning feedback system, cross coupling between feedback loops with and without direct rf feedback etc. As a special case we have shown that the model agrees with the well known Pedersen model derived for the CERN PS booster. As an application of the model we undertook a detailed study of the cross coupling between the loops by considering all of them at once for varying time, Q and beam intensities. A discussion of the method to identify the coupling is shown. At the end a summary of the identified loop interactions is presented.

Application of optimal control theory to optimize the parameters of the low-level rf beam control loops is shown for a low-and a high-intensity circular accelerator. The parameters are: synchronization phase error, beam position error, radial position error, cavity gap voltage error, cavity phase error, cavity tuning error, frequency of the rf system, amplitude of the generator current, phase of the generator current, and tuner bias current. The low-intensity machine is studied by considering the radial, synchronization, and beam phase loops and by ignoring the cavity dynamics. Later we include the cavity model and cover the dynamics of the accelerator system with amplitude, phase, and tuning loops. Flow charts of the computer program are shown to predict and shape the optimal gains starting from the specification on the parameters. The gains are implemented in a particle-tracking code, and with the closed loop system in operation the parameters are tested to be within specification.

In this paper, we present a 3-step procedure to robustly control the revolute flexible-joint manipulator in the presence of parameter variations and bounded input disturbances such as torque ripples. By treating the difference of motor angle and link angle as the input to the rigid link part of the manipulator dynamics, our first step is to design a smooth adaptive reference signal for this input to globally stabilize the rigid subsystem. The second step is to drive the difference of motor and link angles to this desired reference signal exponentially by using sliding control. In the third step we exploit the model reduction capability of sliding control to perform the stability analysis. It is well-known that sliding control can reduce the system order by n if the number of control inputs is n. The exploitation of this property of sliding control makes our stability analysis a lot simpler than other approaches. Global stability in the sense of Lyapunov can be guaranteed and errors in link position and velocity are driven to zero when the system is in sliding mode. No weak elasticity assumption is needed.

A generalized beam control model is derived in terms of accelerator parameters. Using this model, stability conditions are derived for several loop configurations with and without a synchronization loop in the feedback system in the absence of beam-loading effects. A new synchronization scheme analyzed in this paper enables phase-locking of the reference bunch in the lower-energy machine with a reference bucket in the higher-energy machine with controlled phase slippage when ideal phase values are suitably adjusted for transfer. The mathematical technique used here greatly enhances the stability analysis for fast-cycling machines. Analogy with commonly used frequency domain techniques is shown with some examples. Limits under model approximations are presented. Detailed derivations are shown for the configuration planned for the Low Energy Booster of the Superconducting Super Col-lider. Only the results are tabulated for other loop configurations. The approach indicates explicitly how the time-varying gains can be designed for fast-cycling machines. In cases where the derivative of the synchronous phase and the momentum error are large, the method shows how to obtain a suitable compensation.

This paper summarizes a new investigation of applying advanced pansharpening algorithms to enhance the images of the left imager in the Mastcam onboard the Curiosity rover, which landed on Mars in 2012. The various instruments on the rover have already made great contributions in the understanding of Mars. The goal of our research is to generate both high spatial and high spectral image cube by using the left and right Mastcam imagers. Eleven algorithms have been investigated using five objective performance metrics. Subjective evaluations have also been conducted. The image enhancement results are encouraging.

This paper summarizes some preliminary results in enhancing the spatial resolution of the left Mastcam images of the Mars Science Laboratory (MSL) onboard the Mars rover Curiosity. There are two multispectral Mastcam imagers, having 9 bands in each. The left imager has wide field of view, but low resolution whereas the right imager is just the opposite. Our goal is to investigate whether we can use the right Mastcam images to enhance the left Mastcam images. We first estimate the point spread function (PSF) between a pair of left and right Mastcam images using a sparsity based approach. We then apply the estimated PSF to enhance the other left images. Actual Mastcam images were used in our experiments. Preliminary results indicated that the image enhancement performance is mixed. That is, we can achieve good results in some left images and poor results in others. The mixed results point to a new direction for a future study, which involves the use of deep learning based on convolutional neural network (CNN) for PSF estimation and robust deblurring.

This paper presents a novel approach to fusing Thermal Emission Imaging System (THEMIS) and Thermal Emission Spectrometer (TES) satellite images, aiming to improve Mars surface characterization performance from orbit. Our approach includes proven registration and advanced pansharpening algorithms developed by us and others. Preliminary experiments show that the fusion approach is highly promising despite the extremely high resolution difference of THEMIS and TES (30 to 1). We also observed some potential issues that require further research.

Pansharpening refers to the fusion of a high spatial resolution panchromatic image with high spectral resolution multispec-tral or hyperspectral images (MSI or HSI) to yield high resolution data in both spectral and spatial domains. It has been widely adopted as a primary preprocessing step for numerous applications. In this paper, we perform a literature survey of various pansharpening algorithms including the most advanced deep learning approaches for both multispectral and hyperspectral images. We further evaluate the effect of the resolution difference on anomaly detection. Synthetic mul-tispectral and hyperspectral images are generated to evaluate the performance of anomaly detection on high resolution images. Eight state-of-the-art MSI and HSI pansharpening methods are compared in this paper. Experimental results show that, performing anomaly detection on high resolution images improves the detection rate, and at the mean time suppresses the false alarm rate.

The RGBW color filter arrays (CFA) contains R, G, B, and white pixels. It is also known as CFA2.0, which has more luminance information than that of the standard Bayer pattern. A standard demosaicing algorithm was included in the patent associated with CFA2.0. In this paper, we cast the debayering/demosaicing problem for CFA2.0 as a pansharpening problem. The new formulation is very flexible in that it enables us to plug in many pansharpening algorithms for demosaicing. We then compare the standard and the pansharpening approaches. Two benchmark image sets (IMAX and Kodak) were used in our evaluations. It was observed that the standard demosaicing approach worked well for Kodak images, but the pansharpening approaches worked better for the IMAX images. Finally, we also present a comparative study between the standard Bayer pattern and the CFA2.0.

Many remote sensing applications require a high-resolution hyper-spectral image. However, resolutions of most hyperspectral imagers are limited to tens of meters. Existing resolution enhancement techniques either acquire additional multispectral band images or use a pan band image. The former poses hardware challenges, whereas the latter has limited performance. In this paper, we present a new resolution enhancement method that only requires a color image. Our approach integrates two newly developed techniques in the area: (1) A hybrid color mapping algorithm, and (2) A Plug-and-Play algorithm for single image super-resolution. Comprehensive experiments using real hyperspectral images are conducted to validate and evaluate the proposed method.

This paper summarizes some new results in improving the left Mastcam images of the Mars Science Laboratory (MSL) onboard the Mars rover Curiosity. There are two multispectral Mastcam imagers, having 9 bands in each. The left imager has wide field of view, but low resolution whereas the right imager is just the opposite. Our goal is to investigate the possibility of fusing the left and right images to form high spatial resolution and high spectral resolution data cube so that stereo images and data clustering performance can be improved. Many pansharpening algorithms have been investigated. Actual Mastcam images were used in our experiments. Preliminary results indicate that the pansharpened images can indeed enhance the data clustering performance using both objective and subjective evaluations.

In this research, we developed an adaptive filtering system for enhancing auscultation performance in noisy environments such as spacecraft and International Space Station (ISS). The system uses a stethoscope, a microphone, and an adaptive filter. Four filtering algorithms (least mean square (LMS), normalized LMS (NLMS), recursive least square (RLS), and our patented algorithm known as Frequency-domain Minimum Square Error with length N and Signal Detection (FMSENSD)) were implemented and compared. Extensive experiments using actual data collected by several commercial stethoscopes clearly demonstrated the performance of the system under noisy conditions up to 79 dBA.

This paper summarizes some preliminary results of applying deep belief network (DBN) to land classification using hyperspectral images. The performance of DBN is then compared with several conventional classification approaches. A fusion approach is also proposed to combine spatial and spectral information in the classification process. Actual hyperspectral image data were used in our investigations. Based on the particular data and experiments, it was found that DBN has slightly better classification performance if only spectral information is used and has slightly inferior performance than a conventional method if both spatial and spectral information are used.

— This paper describes a novel approach to sensor and actuator integrity monitoring. Multiple sensor and actuator faults can be detected and isolated. Most importantly, fault magnitudes can be correctly estimated. Our approach is robust to disturbance and does not require additional sensors.

— This paper summarizes our research results on local active noise reduction. Our aim is to create a small quiet zone for astronauts in noisy spacecraft environments. A novel approach is proposed. Extensive simulations and preliminary real-time experiments demonstrate the efficacy of the proposed system.

Anomaly detection has been known to be a challenging, ill-posed problem due to the uncertainty of anomaly and the interference of noise. In this paper, we propose a novel low rank anomaly detection algorithm in hyperspectral images (HSI), where three components are involved. First, due to the highly mixed nature of pixels in HSI, instead of using the raw pixel directly for anomaly detection, the proposed algorithm applies spectral unmixing algorithms to obtain the abundance vectors and uses these vectors for anomaly detection. Second , for better classification, a dictionary is built based on the mean-shift clustering of the abundance vectors to better represent the highly-correlated background and the sparse anomaly. Finally, a low-rank matrix decomposition is proposed to encourage the sparse coefficients of the dictionary to be low-rank, and the residual matrix to be sparse. Anomalies can then be extracted by summing up the columns of the residual matrix. The proposed algorithm is evaluated on both synthetic and real datasets. Experimental results show that the proposed approach constantly achieves high detection rate while maintaining low false alarm rate regardless of the type of images tested.

Anomaly detection becomes increasingly important in hyper-spectral image analysis, since it can now uncover many material substances which were previously unresolved by multi-spectral sensors. In this paper, we propose a Low-rank Tensor Decomposition based anomaly Detection (LTDD) algorithm for Hyperspectral Imagery. The HSI data cube is first mod-eled as a dense low-rank tensor plus a sparse tensor. Based on the obtained low-rank tensor, LTDD further decomposes the low-rank tensor using Tucker decomposition to extract the core tensor which is treated as the " support " of the anomaly spectral signatures. LTDD then adopts an unmixing approach to the reconstructed core tensor for anomaly detection. The experiments based on both simulated and real hyperspectral data sets verify the effectiveness of our algorithm.

Traditional hyperspectral anomaly detection methods either model the global background or the local neighborhood, that bring some apparent drawbacks, such as the unreasonable assumption of uni-modular background in global detectors, or the high false alarms by sliding windows in local detectors. In this paper, a source component-based anomaly detection approach is proposed. It first extracts the source components in the spectral image data cube by using the blind source component separation and then identifies the components that are anomaly (or salient) to other components. We interpret the anomaly detection as a matrix decomposition problem with the minimum volume constraint for the multi-modular background and sparsity constraint for the anomaly image pixels. Experimental results show that the approach is promising for anomaly detection in spectral data cube.

JMARS (Java Mission-planning
and Analysis for Remote Sensing) is a geospatial information
system (GIS) developed by ASU's Mars
Space Flight Facility to provide mission planning and
data analysis tools for NASA planetary mission data to
scientists, students of all ages, and to the general public
[1]. We developed a custom layer for JMARS to show
the traverse map of Mars rovers including Spirit, Opportunity
and Curiosity (see Fig. 1). The tool allows
users to easily view spectral measurements obtained by
the rovers (Fig. 2) and concentration results (Fig. 4)
generated by scientists. When a particular sol (Mars
day) is selected, the graphics window of the JMARS
software shows the location of the rover at that day
(see Fig. 3). In Fig. 1 and Fig. 3, we also load HiRISE
data layer to show the high resolution image of Mars.

Compositional analysis is important to interrogate spectral samples for direct analysis of materials in agriculture, environment and archaeology, etc. In this paper, multi-variate analysis (MVA) techniques are coupled with laser induced breakdown spectroscopy (LIBS) to estimate quantitative elemental compositions and determine the type of the sample. In particular, we present a new multivariate analysis method for composition analysis, referred to as "spectral unmixing". The LIBS spectrum of a testing sample is considered as a linear mixture with more than one constituent signatures that correspond to various chemical elements. The signature library is derived from regression analysis using training samples or is manually set up with the information from an elemental LIBS spectral database. A calibration step is used to make all the signatures in library to be homogeneous with the testing sample so as to avoid inhomogeneous signatures that might be caused by different sampling conditions. To demonstrate the feasibility of the proposed method, we compare it with the traditional partial least squares (PLS) method and the univariate method using a standard soil data set with elemental concentration measured a priori. The experimental results show that the proposed method holds great potential for reliable and effective elemental concentration estimation.

The ChemCam instrument package on the Mars rover, “Curiosity”, is the first planetary instrument that employs laser induced breakdown spectroscopy (LIBS) to determine the compositions of geological samples on another planet. However, the sampled spectra are often corrupted by various sources of interferences that would largely affect the accuracy of elemental concentration estimation. Therefore, preprocessing is essential to improve the quality of the spectra. This paper revisits the conventional preprocessing procedures where denoising is followed by continuum removal. Through comprehensive performance evaluation, we propose a new procedure that would lead to much improved estimation accuracy. First, we show that the denoising process should be conducted after continuum removal. Second, a state-of-the-art image denoising technique is adapted to the 1D domain to boost the performance of denoising. Third, an additional preprocessing step is added that effectively select the most info...

JMARS (Java Mission-planning
and Analysis for Remote Sensing) is a geospatial information
system (GIS) developed by ASU's Mars
Space Flight Facility to provide mission planning and
data analysis tools for NASA planetary mission data to
scientists, students of all ages, and to the general public
[1]. It provides convenient accessing and visualization
of imagery data coming from various Mars missions
such as MGS TES, Odyssey THEMIS, MER MiniTES,
and MRO HiRISE. Map and image datasets
from Mars missions are generally at the size of terabytes.
JMARS breaks each file into small pieces and
stores them in the database so data computation and
images can be produced quickly.
Chemical composition estimation is one of the important
goals of past, present and future Mars rovers.
The existing Spirit and Opportunity rovers are
equipped with millimeter wave imaging spectrometers
to detect the chemical composition in Mars rocks.
Each time a rover comes close to a rock, it removes the
Martian dust from the surface and takes the images of
the rock. The probe continues to carry out the same
procedure at different locations of the same rocks in
order to collect enough spectral information to extract
the features and classify the chemical composition of
the rock. It is a time-consuming process to identify a
special location which may contain certain abnormal
chemical composition in the rock. Recently, NASA
scientists found that Laser Induced Breakdown Spectroscopy
(LIBS) and Alpha Particle X-ray Spectrometer
(APXS) techniques are able to carry out the analysis
of the rock’s chemical composition on Mars with much
better performance. LIBS can detect the chemical
composition in a fast pace and APXS can identify the
composition with much higher accuracy.
We are developing a Java version of chemical
composition detection tool for JMARS package to enhance
the current software version, allowing users to
examine and analyze collected LIBS spectra. The
sample spectra would be processed through a chemical
composition estimation system to obtain a list of elemental
compositions of the target rocks. Similarly, for
any collected APXS spectrum of the rocks or soil samples,
the tool will produce a list of elemental composition
with more accurate percentages. The soil or rock
samples with different characteristics as compared to
earlier collections of soil/rock samples will also be
detected with this tool. The generated results will then
be sent back to JMARS users for display.

NASA is developing a new generation of audio system for astronauts. The idea is to use directional speakers and microphone arrays. However, since the helmet environment is very reverberant, the inbound signals in the directional speaker may still enter the outbound path (microphone array), resulting in an annoying positive feedback loop. To improve the communication quality between astronauts, it is necessary to develop a digital filtering system to minimize the interactions between inbound and outbound signals. In this paper, we will present the following results. First, we set up experiments under three scenarios: office, bowl, and helmet. Experiments were then performed. Second, 3 adaptive filters known as normalized least mean square (NLMS), affine projection (AP), and recursive least square (RLS) were applied to the experimental data. We also developed a new frequency domain adaptive filter called FDAFSS (frequency domain adaptive filter (FDAF) with spectral subtraction (SS)), which is a combination of FDAF and SS. FDAFSS was compared with LMS, AP, RLS, FDAF, and SS filters and FDAFSS yielded better performance in terms of perceptual speech quality (PESQ). Moreover, FDAFSS is fast and can yield uniform convergence across different frequency bands.

Develop high performance and real-time processing tools for
processing HyspIRI data (VSWIR spectrometer and thermal infrared
(TIR) imager) to enhance material identification and change detection
applications.
• Demonstrate the feasibility of the approach using realistic data.

Currently, rock sample selection for Mars exploration is done manually. Human operators on Earth provide remote guidance on which rock sample to select and rovers follow the order. Due to long communication delay between rovers and human operators, this process is very inefficient. In the new Mars rover, Curiosity, laser induced breakdown spectroscopy (LIBS) is an important rock sample collection instrument. LIBS can provide fast and quick sample collection and hence it can be used as a prescreening tool. However, an automatic rock sample analysis software tool is needed. In this paper, we present two software tools to analyze LIBS data. One is the well known partial least square (PLS) method and the other one is known as nonnegatively constrained least square (NCLS) method. Since no real Mars LIBS samples are available at this moment, we used LIBS data from gold alloy to evaluate the above two techniques. We observed that PLS provides better concentration estimates than NCLS for gold alloy.

Figure 1: Compression scheme defined by the ratio between compressed size and un-compressed size, i.e. For lossy compression, there is a tradeoff between compression ratio and the result quality. The higher the quality is, the larger the ratio will be. A common quality measurement is peak signal to noise ratio (PSNR) Abstract Data compression is one of challenging problems in data communication system due to the information explosion. In this paper, we propose a novel and comprehensive compressive sensing-based system for data compression. Performance of our proposed system is compared with conventional compression algorithm such as Huffman coding in terms of mean square error (MSE) after decompression, computation complexity and compression ratio, etc. As an application example, we implement this system to real world wind tunnel data. Simulation results show that our system can yield comparable or even better compression as Huffman coding in terms of information loss. The major drawback of Huffman coding is to calculate the probability of each symbol which means it is not be appropriate for real time coding due to large amount of calculation. Meanwhile, our proposed system can process data by multiplying original data with Gaus-sian or Bernoulli sensing matrix directly which is also easy to implement. MAX PSNR = 201og10[RMSE] (2)

In this paper, we summarize our efforts of using three different radars (impulse radar, swept frequency radar, and continuous-wave radar) for through-the-wall sensing. The purpose is to understand the pros and cons of each of the three radars. Through extensive experiments, it was found that the radars are complementary and multiple radars are needed for different scenarios of through-the-wall target detection and tracking.

 The integrity of underground power distribution networks is
an important and challenging topic.
 Primary challenges:
1. Multiple power injection points.
2. Many loads, including capacitive and inductive loads
that absorb some the fault transients.
3. The fault currents in incipient stages are small.
 In this study, we propose a low-cost, fast and accurate
approach to localize short-circuit faults in underground
power networks

Con Edison experiences more than 1000 arcing faults on its secondary distribution system each year. Arcing faults introduce strong harmonics into the power network. We propose a fast, low cost, and high performance approach to localizing arcing faults based on harmonics. First, we implemented 2 novel algorithms. One is based on voltage ratio of harmonics. This method can detect arcing fault one at a time. The second one is based on sparse sensing, which is a powerful technique that can detect multiple faults and is robust to measurement noise. Both methods require low cost voltage measurements; no high cost sensors are required. In addition, the computations can be done very quickly within a few cycles (

Hyperspectral imaging is becoming more and more popular in military surveillance reconnaissance operations. However, one serious limitation of hyperspectral images is that many constituent members (materials) may be present in a single pixel, making accurate target detection or anomaly detection extremely challenging. In this paper, we present a comparative study of several recently developed, very effective unsupervised unmixing algorithms to detect anomaly with sub-pixel accuracy. The unsupervised algorithms do not assume any prior knowledge of the target signature and can be readily deployed in real-world applications. The algorithms include minimum volume constrained non-negative matrix factorization (MVCNMF), gradient descent maximum entropy (GDME) and unsupervised fully constrained least squares (UFCLS). Actual hyperspectral image containing 4 panels and 2 small targets from the Air Force is used in our studies. The experimental results show that MVCNMF achieves the best detection results and UFCLS presents close performance to MVCNMF but better results than GDME. In addition, the speed of MVCNMF is the fastest among the three methods. We also observe that some simple dimensionality reduction algorithms, such as principal component analysis (PCA) applied to reduce the dimensionality of the raw data before the unmixing operation, would affect the detection performance in a negative way.

SUMMARY: Computational speed and performance are very important for real-time reconnaissance and surveillance using hyperspectral images. Here we present a few new and fast algorithms. First, we propose a new fast algorithm called cluster kernel RX. Our methods can significantly improve the computational speed without sacrificing the anomaly and change detection performance. Second, several prediction methods using neural network (NN), local information, and multiple references are also proposed. The methods have great potential in dealing with misregistration and parallax issues. Evaluation using actual hyperspectral images from the Air Force showed that the above algorithms are promising.

Endmember extraction in Hyperspectral Images (HSI) is a critical step for target detection and abundance estimation. In this paper, we propose a new approach to endmember extraction, which takes advantage of the sparsity property of the linear representation of HSI's spectral vector. Sparsity is measured by the l 0 norm of the abundance vector. It is also well known that l 1 norm well resembles l 0 in boosting sparsity while keeping the minimization problem convex and tractable. By adding the l 1 norm term to the objective function, we result in a constrained quadratic programming which can be solved effectively using the Linear Complementary Programming (LCP). Unlike existing methods which require expensive computations in each iteration, LCP only requires pivoting steps, which are extremely simple and efficient for the un-mixing problem, since the number of signatures in the reconstructing basis is reasonably small. Preliminary experiments of the proposed methods for both supervised and unsupervised abundance decomposition showed competitive results as compared to LS-based method like Fully Constrained Least Square (FCLS). Furthermore, combination of our unsupervised decomposition with anomaly detection makes a decent target detection algorithm as compared to methods which require prior information of target and background signatures.

SUMMARY: In target detection, it is normally assumed that the ground truth signature, collected in laboratory environment, of a target is available and we can then use it to search for targets in a given hyperspectral image. However, directly applying the ground truth signature to the real data is not appropriate due to the environmental differences between the ground truth data and real data. The environmental factors include, but not limited to, illumination, seasonal changes, temperature differences, and sensor characteristics (sensor in laboratory and sensor in use). By compensating for the environmental factors, the target detection performance could be improved significantly. This paper summarizes our efforts on a new practical approach to high performance target detection in hyperspectral images. We present results of two compensation techniques. The first one is ISAC (in-scene atmospheric correction) [1] [2] and the second one is a Hybrid ISAC (H-ISAC) method developed by us. Based on our simulations, we found that ISAC does not work for the hyperspectral images from the Air Force. The reason is that ISAC assumes a blackbody model, which only applies to wavelengths longer than 1 m µ. However, the wavelengths of the Air Force images are smaller than 1 m µ , which makes the blackbody assumption invalid. Our H-ISAC method uses a training image to establish a mapping between ground truth signature domain and testing image domain. Simulation results showed that the H-ISAC algorithm provides excellent compensation to environmental effects. After compensation, the performance of target detection using receiver operating characteristics (ROC) is significantly improved.. In this paper, a novel compensation scheme called hybrid ISAC (H-ISAC) is proposed. Similar to in-scene vegetation normalization scheme, we use in-scene vegetation information to estimate w. However,

—Rotor imbalance in induction machines has been widely studied. The signatures to look for in the stator current for detecting rotor imbalance are well established. However, an accurate explanation for the appearance of these signatures is lacking. Moreover, in most studies only a single phase of stator current has been used for detecting rotor imbalance and determining its severity. Combing imbalance features from the three phases through sensor fusion can yield more accurate and reliable results. Therefore, this paper focuses on, (i) accurate modeling of rotor imbalance to explain the genesis of its signatures in the stator current and, (ii) determining imbalance severity by sensor fusion. A test bed is established to verify the proposed approach. Index Terms— Adaptive Network based Fuzzy Inference System (ANFIS), Dempster Shafer (DS) Theory, Fuzzy c means Clustering, Space Phasor Theory, Squirrel Cage Motors.

— In this paper, we summarize our research activities in Condition Based Maintenance (CBM) of critical power system components using Wireless Sensor Network (WSN). First, two testbeds were built: one for emulating electrical faults in motor windings and one for emulating mechanical faults in motors and generators. Second, appropriate sensors were installed on the testbeds and sensor data were collected using wireless nodes. Third, advanced algorithms were implemented and extensive simulations validated the performance of the algorithms. Finally, real-time experiments were performed to detect various faults.

—Conventional speaker identification and speech recognition algorithms do not perform well if there are multiple speakers in the background. For high performance speaker identification and speech recognition applications in multiple speaker environments, a speech separation stage is essential. Here we summarize the implementation of three speech separation techniques. Advantages and disadvantages of each method are highlighted, as no single method can work under all situations. Stand-alone software prototypes for these methods have been developed and evaluated.

—Here we summarize the application of Hidden Markov Model (HMM) to bearing diagnostics. Our proposed HMM incorporates fault diagnostics and fault severity estimation in an integrated framework. Actual bearing data were used to demonstrate the performance of the framework.

—Noisy environments seriously degrade the performance of speech recognition systems. Here we implement a high performance speech enhancement algorithm. Data from Speech Separation Challenge [1] were used to evaluate the method. It was observed that the enhanced speech significantly improved the recognition performance. In 2 out of 4 SNR cases, over 100% relative percentage improvements were achieved. Standalone software prototype has been developed and evaluated.

Low cost and portable gas chromatography-ion mobility spectrometry (GCIMS)
has been used to distinguish some toxic chemicals
• Challenges:
- The concentration of chemicals may be very low and it may be difficult to
determine the retention time (RT) and drift time of the chemicals.
- Chemicals with different concentrations tend to have different
characteristics, especially in the IMS drift tube. For example, at low
concentrations, chemicals seldom form dimers, which are heavier due to the
formation of a large mass between water and the chemical. Monomers and
dimers may confuse the chemical classification process.
- It is important to estimate the concentration of chemicals, as this
information will provide toxicity. However, low concentration chemicals are
extremely difficult to detect and estimate.

—Conventional speaker identification and speech recognition algorithms cannot deal with noisy and multiple speaker environments. For example, IBM via Voice has low recognition rates if dictation is done in a noisy environment. In order to achieve high performance in speaker identification and speech recognition, we propose an integrated approach that takes every facet of the process into account. Here we summarize some preliminary results from the application of this integrated approach to robust speaker identification and speech recognition. A real-time stand-alone software prototype has been developed to evaluate the effectiveness of the approach.

— Application results of an adaptive prognostic approach to actual bearing data are summarized in this paper. The on-line learning capability and no historical data requirement are two advantages of the investigated approach. Because the lifetime model parameters are updated on-line using the latest diagnostic information, the remaining useful life (RUL) prediction performance is highly accurate. The approach could be applied in real-time which makes it feasible for use in industry. Actual bearing data were used to demonstrate the efficacy of the approach. Confidence intervals for RUL predictions have been computed.

In this paper, we summarize our research activities in Condition Based Maintenance (CBM) of critical power system components using Wireless Sensor Network (WSN). First, two testbeds were built: one for emulating electrical faults in motor windings and one for emulating mechanical faults in motors and generators. Second, appropriate sensors were installed on the testbeds and sensor data were collected using wireless nodes. Third, advanced algorithms were implemented and extensive simulations validated the performance of the algorithms. Finally, real-time experiments were performed to detect various faults.