COLLEGE RECRUITING
2007–2008 Technical Seminar Series
College Recruiting Program Administrator
Human Resources
Lincoln Laboratory
Massachusetts Institute of Technology
244 Wood Street
Lexington, MA 02420-9108
(781) 981-2465
email: collegerecr@ll.mit.edu
Correlating Sensor Tracks with the Deterministic Annealing Method
Brian M. Lewis , PhD |
Matthew A. Horsley, PhD |
Surveillance systems often involve multiple sensors tracking multiple
objects. To maximize the preservation of information gained by each sensor, a mapping must be performed between the objects in each sensor’s field of view. Due to differing sensitivity and geometry, the number of objects detected by each sensor might not be the same. Object correlation is further complicated because there tends to be sensor bias due to pointing errors. This talk discusses the deterministic annealing approach to object correlation and sensor bias removal.
Advanced Signal Processing and Three-Dimensional Imaging Techniques for Wideband Radars
Joseph T. Mayhan, PhD
Electrical Engineering
The Ohio State University
MIT Lincoln Laboratory has led the nation in the development of high-power wideband radar technology. Such radars have a unique capability for resolving scattering centers and producing three-dimensional images of individual targets. In parallel with developing these advanced radar systems, MIT Lincoln Laboratory has developed high-speed signal and data processing techniques and algorithms that permit the generation of target images and extraction of target features in near real time. This seminar will present an overview of these high-power sensors and associated processing techniques, including such techniques as ultra-wideband, sparse band, coherent processing, and three-dimensional “snapshot” imaging and feature extraction.
The Use of the Gauss-Markov Theorem in Winds Analysis
Rodney E. Cole, PhD
Mathematics
University of Colorado
Understanding the wind is important to many sectors of society. To better measure winds, the National Weather Service and the Federal Aviation Administration (FAA) have deployed numerous Doppler weather radars around the country, as well as many ground stations, and airlines routinely downlink winds while in flight. The best picture of the winds emerges if these many different data sources can be combined into a single coherent wind field. A number of factors complicate the joint analysis of these different data. The data have differing accuracy and a nonuniform distribution. The Doppler data are especially problematic in that each Doppler return provides only a single component of the wind vector.
The Gauss-Markov Theorem provides a framework for a least-squares fit of the wind field to the data that takes into account sensor error and error correlations due to the nonuniform data distribution. Like three-dimensional variational techniques, the measurements do not need to be made in model variables, so the Doppler data can be used without ad hoc transformations into wind vectors. The Gauss-Markov approach can be shown to include traditional optimal interpolation (OI) methods, without some of the limitations of OI. The winds analysis developed for the FAA Integrated Terminal Weather System uses this technique and is the first operational real-time winds analysis system to utilize multi-Doppler techniques.
Aircraft Separation Standards and Radar Performance in the National Airspace System
Steven D. Thompson, PhD
Nuclear Engineering
Georgia Institute of Technology
To maintain safety in the Air Traffic Control system, the Federal Aviation Administration (FAA) establishes the minimum distances between which air traffic controllers must separate aircraft. These distances directly impact the capacity and overall system delay of the Air Traffic Control system and are based, in part, on the performance of the surveillance systems used in tracking aircraft. Today’s modern air traffic control radar systems rely on both “primary” reflections from the aircraft itself and onboard devices designed to electronically reply to a radar sensor’s “interrogation” signal. Aircraft tracking based on the replies from an aircraft’s transponder is known as “secondary” surveillance and is the dominant form of tracking and providing separation in the air traffic control system today.
The minimum separation between aircraft using radar surveillance has been three nautical miles for any sensor and required that both aircraft being separated be tracked by the same sensor and be within 40 nautical miles of that sensor. If these requirements are not met, the aircraft must be separated by five nautical miles or more with a corresponding impact on airspace capacity.
This presentation reviews the separation standards and their origins and discusses the different surveillance environments in use today in the National Airspace System. The error sources and characteristics of the sensors currently used to track aircraft are reviewed and discussed, and a metric is derived to facilitate comparison of the performance of various sensor designs used in different surveillance environments. It was found that modern “monopulse” radar sensors offer a significant performance advantage over older designs and the potential for expanded use of three mile separation, both in the single-sensor environment and in mosaic environments involving the use of multiple sensors. As a result of this analysis, the FAA has recently extended the use of three-mile separation for aircraft within 60 nautical miles of monopulse sensors. A practical application of the results of this analysis is presented in a case study for three-mile separation within Boston’s Air Route Traffic Control Center.
Integrating Unmanned Aerial Vehicles Safely into the National Airspace System
Mykel J. Kochenderfer, PhD
Informatics
University of Edinburgh
Unmanned aerial vehicles (UAVs) such as the Air Force’s Global Hawk and Predator are increasingly employed by the military in roles that require sharing airspace with civilian aircraft. Many civil applications of UAVs have also been proposed for tasks that include border patrols, highway and agricultural observation, and cargo transport. Due to the pressure for widespread access for UAVs and the risk of collision with passenger aircraft, the U.S. is rapidly facing serious safety concerns.
This seminar provides an overview of the safety issues of UAVs sharing airspace with passenger airplanes, including methods for evaluating and mitigating collision risk. The presentation will describe an initiative at MIT Lincoln Laboratory to assess the safety of one particular application, the Traffic Alert and Collision Avoidance System (TCAS), on the Global Hawk UAV. TCAS is in use on all large passenger-carrying aircraft as an independent, last-resort collision avoidance system. TCAS uses a set of sensors and algorithms to detect potential mid-air collisions, alert the flight crew, and provide guidance so that they can avoid a collision. Adapting TCAS to Global Hawk, which has unconventional flight characteristics and a pilot who may be thousands of miles away, creates challenges that require a thorough safety analysis before TCAS might be accepted by domestic and international communities.
This seminar describes the overall approach to safety analysis, including the use of fast-time simulation of UAVs and conventional air traffic. Finally, extensions to future applications, such as onboard infrared, visual, or radar detection systems designed to “see and avoid” other aircraft, will be discussed.
Network Flow Decision-Making Under Uncertainty in the National Airspace System
James E. Evans, PhD
Electrical Engineering
Massachusetts Institute of Technology
Most of the rapid increase in U.S. aviation delays over the past six years has occurred in months characterized by thunderstorm activity. Thunderstorms present a very difficult system management challenge because
- both terminal and en route capacities are significantly reduced by phenomena that are difficult to predict in advance, and
- the dynamic response of the air route network to possible changes in traffic flows is poorly understood.
An important issue in addressing this problem is the trade-off between predictability and optimality of the control actions used. In particular, supplying corrective actions such as delaying or rerouting aircraft hours in advance provides the greatest opportunities for achieving an optimal response if the capacity impacts are accurately known. However, where there is major uncertainty as to the nature of the capacity loss, implementing controls early may be very nonoptimal.
In this talk, we discuss the early stages of a new paradigm of dynamically revised plans that start with an initial multi-hour plan based on inaccurate longer lead-time forecasts, but then use rapidly updated weather forecasts and three-dimensional storm information coupled with air traffic management decision tools that help Federal Aviation Administration air traffic control and airline operations center personnel rapidly develop and implement shorter lead-time weather-impact mitigation plans. Recent operational experience with an experimental system in the Great Lakes and Northeast corridors will be presented.
The talk will conclude with a discussion of opportunities for creative contributions by university researchers in areas such as stochastic optimization for aircraft routing and traffic flow optimization, probabilistic weather modeling, and delay causality analysis.
Multifunction Phased-Array Radar for U.S. Civil-Sector Surveillance Needs
Mark E. Weber, PhD
Physics
Rice University
This seminar describes a concept study for possible future utilization of active electronically scanned (AES) radars to provide weather and aircraft surveillance functions in the U.S. airspace. If transmit-receive element costs decrease sufficiently, multifunction AES radars might prove to be a cost-effective alternative to current surveillance radars (WSR-88D, TDWR, ASR, ARSR) since the number of required radars would be reduced and maintenance and logistics infrastructure would be consolidated. A radar configuration that provides terminal-area and long-range aircraft surveillance and weather measurement capability is described, and a radar network design that replicates or exceeds current airspace coverage is presented. Key technology issues are examined, including transmit-receive elements, overlapped subarrays, the digital beamformer, and weather and aircraft postprocessing algorithms. The presentation concludes with a discussion of how such a radar network might integrate with next-generation weather and aircraft surveillance architectures.
Disease Modeling to Assess Outbreak Detection and ResponseDiane C. Jamrog, PhD |
Bioterrorism is a serious threat that has become widely recognized since the anthrax mailings of 2001. In response, one national research activity has been the development of biosensors and networks thereof. A driving factor behind biosensor development is the potential to provide early detection of a biological attack, thereby enabling timely treatment. This presentation introduces a disease progression and treatment model to quantify the potential benefit of early detection. To date, the model has been used to assess responses to inhalation anthrax and smallpox outbreaks.
Defending Against Biological Terrorism
Timothy J. Dasey, PhD
Biomedical Engineering
Rutgers University
The intentional use of disease as a weapon has been recorded throughout human history. With modern technology, the potential impact of biological attacks is exceeded by only nuclear weapons. The dramatic advances in the biological sciences in the last few decades, and the expertise accumulated in large state-funded offensive biological weapons programs, have made the process simpler for those who are inclined to pursue such weapons.
The U.S. is in the midst of an aggressive research and development program to prevent, detect, and mitigate the effects of biological attacks. Research thrusts include rapid detection and characterization of biological agents for surveillance, response, or forensic purposes. Strategies for organizing and deploying medical resources following an attack are being developed and tested. Improved medicines are needed for prophylactic and therapeutic purposes. Information fusion and data mining techniques are needed to intelligently process diverse data sets for detection and response to an attack. Modeling and simulation of attacks and remedies are used to develop system requirements and designs and to test defensive implementations. Extensive environmental and clinical measurements are needed to understand the performance of technologies and for proper parameterization of simulations.
This presentation will frame the problem by describing the important characteristics of biological weapons, discussing the technical challenges in defending against these weapons, and providing examples of the technologies and systems being developed at MIT Lincoln Laboratory to address this problem. The wide breadth of research disciplines in this field will be highlighted.
Adaptive SAR Results with the LiMIT Testbed
Gerald R. Benitz, PhD
Electrical Engineering
University of Wisconsin–Madison
Adaptive phased-array radars promise significant performance enhancements for synthetic aperture radar (SAR). One obvious application is to provide electronic protection (EP) against interference. Another application is to mitigate range and/or Doppler ambiguities, thereby potentially enabling multiple simultaneous modes of operation, increased area rates, and higher-resolution scan modes.
Presented here are adaptive processing results from the Lincoln Multi-mission ISR Testbed (LiMIT) developed by MIT Lincoln Laboratory in 2004. The system is an 8-channel X-band radar installed on a Boeing 707. Initial flight tests employed a 180 MHz waveform and digitally recorded spotlight SAR mode collections. Collections at Ft. Huachuca in July 2004 included wideband interference sources. EP results demonstrate SAR image recovery via adaptive processing, with improvements in signal-to-interference ratio exceeding 30 dB for a near-main-beam interferer. Mitigation approaches include derivative-based updating in the frequency domain and multiple time-taps in the image domain. The impact on image quality is minimal. Doppler ambiguity mitigation is demonstrated by producing an image from the transmit beam sidelobes. The sidelobe region has Doppler frequencies that are ambiguous with the main lobe region, and only adaptive beam steering can recover this energy. One image result shows recovery of a scene with 6 kHz clutter bandwidth, but from a data collection using only a 2 kHz pulse rate.
GPS Space-Time Adaptive Array Processor Test Results
Gary F. Hatke, PhD
Electrical Engineering
Princeton University
GPS jamming has been acknowledged as a severe threat to military GPS usage during combat, with the number of systems linked to GPS growing daily. It is clear that an effective system for mitigating GPS interference (intentional or unintentional) is needed on some high-value assets. Spatial nulling has been proposed as an effective method for combating GPS jamming, and, in fact, there are available GPS processing units that can provide spatial nulling capabilities when coupled with a 7-element antenna array. For stressing interference environments, however, these spatial-only techniques will have difficulty providing adequate anti-jam (A/J) protection for GPS operations. For this reason, space-time adaptive (STAP) beamforming processors have been proposed for GPS A/J. This talk will introduce the techniques involved in developing an effective STAP beamformer for GPS. Specifically, the design and test performance of one such STAP beamformer, the Multipath Adaptive Multi-Beam Array (MAMBA) processor, will be examined. The design of the processor will be reviewed, and then laboratory test results indicating the expected levels of performance achievable with the MAMBA will be presented. These tests include illuminating the antenna array with multiple broadband jamming signals (along with a GPS signal) to determine the increase in signal-to-interference-plus-noise (SINR) obtained on the GPS signal by MAMBA processing.
Polynomial Rooting Techniques for Adaptive Array Direction Finding
Gary F. Hatke, PhD
Electrical Engineering
Princeton University
Array processing has many applications in modern communications, radar, and sonar systems. Array processing is used when a signal in space, be it electromagnetic or acoustic, has some spatial coherence properties which can be exploited (such as far-field plane wave properties). The array can be used to sense the orientation of the plane wave, and thus deduce the angular direction to the source. Adaptive array processing is used when there exists an environment of many signals from unknown directions as well as noise with unknown spatial distribution. Under these circumstances, classical Fourier analysis of the spatial correlations from an array data snapshot (the data seen at one instance in time) is insufficient to localize the signal sources.
In estimating the signal directions, most adaptive algorithms require computing an optimization metric over all possible source directions and searching for a maximum. When the array is multidimensional (e.g., planar) this search can become computationally expensive, as the source direction parameters are now also multidimensional. In the special case of one-dimensional (line) arrays, this search procedure can be replaced by solving a polynomial equation, where the roots of the polynomial correspond to estimates of the signal directions. This technique had not been extended to multidimensional arrays because these arrays naturally generated a polynomial in multiple variables, which does not have discrete roots.
This talk introduces a method for generalizing the rooting technique to multidimensional arrays by generating multiple optimization polynomials corresponding to the source estimation problem and finding a set of simultaneous solutions to these equations, which contain source location information. It is shown that the variance of this new class of estimators is equal to that of the search techniques they supplant. In addition, for sources spaced closer than a Rayleigh beamwidth, the resolution properties of the new polynomial algorithms are shown to be better than those of the search technique algorithms.
Acoustic Source Depth Discrimination Performance Prediction Analysis| Christ D. Richmond, PhD Electrical Engineering Massachusetts Institute of Technology |
Shawn Kraut, PhD Physics University of Colorado |
| William H. Payne, BSEE Electrical Engineering University of Maryland |
Vitaly Kmelnitsky, MS |
[1] V. Premus, “Normal Mode Filtering for Acoustic Source Depth Discrimination,” Technical Memorandum VEP-ONR-01-2003, MIT Lincoln Laboratory, September 2003.
[2] V. Premus, J. Ward, and C. D. Richmond, “Mode Filtering Approaches to Acoustic Source Depth Discrimination,” Proceedings of the Thirty-Eighth Asilomar Conference on Signals, Systems and Computers, Vol. 2, pp. 1415–1420, November 2004.
Mechanical Systems Engineering of Optical Sensors
Steven E. Forman, PhD
Mechanical Engineering
Harvard University
During the past fifteen years, MIT Lincoln Laboratory has developed several different optical sensor experiments that have flown on airborne and space platforms. These include the Space-Based Visible, Airborne Infrared Imager, and Advanced Land Imager. Each represents a one-of-a-kind sensor fully engineered at MIT Lincoln Laboratory. This talk summarizes several of the mechanical systems engineering areas and issues that occurred throughout design, analysis, fabrication, integration, and testing of these systems. Included are discussions of optical, optomechanical, structural, and thermal engineering; electronic packaging; mechanism design; focal-plane packaging; control-system engineering; materials selection and testing; environmental testing; failure analysis; and computer-aided design and analysis tools.
Integrated Optics in Silicon| Steven J. Spector, PhD Physics SUNY-Stony Brook |
Michael W. Geis, PhD Physics Rice University |
Theodore M. Lyszczarz, PhD
Electrical Engineering
Massachusetts Institute of Technology
Silicon is an attractive platform for the integration of optical components because of the mature infrastructure dedicated to the fabrication of silicon microelectronic devices. Furthermore, the silicon optical components can be readily integrated with transistors to form complex signal processing systems. At the wavelengths most relevant for optical communication (near 1550 nm), silicon has a high index and low inherent loss, making it an ideal material for compact devices. However, such compact devices also have large scattering loss. In addition, the electro-optic effects are weak in silicon, making it difficult to realize active components. This talk will describe techniques, such as smoothing and selective waveguide widening, for reducing scattering loss to enable the fabrication of high-Q filters and low-loss waveguides. The seminar will discuss three types of optical modulators, based on thermal index modulation in a silicon waveguide, carrier-based index modulation in a PIN diode waveguide structure, and a silicon-compatible electro-active organic crystal waveguide.
Optical Sampling for High-Speed, High-Resolution Analog-to-Digital Conversion| Jonathan C. Twichell, PhD Nuclear Engineering University of Wisconsin-Madison |
Paul W. Juodawlkis, PhD Electrical Engineering Georgia Institute of Technology |
The performance of digital receivers used in modern radar, communication, and surveillance systems is often limited by the performance of the analog-to-digital converter (ADC) used to digitize the received signal. Optically sampled ADCs, which combine optical sampling with electronic quantization, have been demonstrated to extend the performance of electronic ADCs. The primary advantages of using optics to perform the sampling function include (1) the timing jitter of modern mode-locked lasers is more than an order of magnitude smaller than that of electronic sampling circuitry, (2) the low dispersion of optical components allows picosecond sampling pulses to be used to attain wide analog bandwidth, and (3) demultiplexing to arrays of time-interleaved electronic converters can be performed in the optical domain rather than in the electrical domain with no signal bandwidth, nonlinearity, or memory effect constraints. MIT Lincoln Laboratory’s work in this area has focused on the development of a linear sampling technique referred to as phase-encoded optical sampling. The technique uses a dual-output Mach-Zehnder electro-optic modulator as a sampling transducer to achieve both high linearity and 60 dB suppression of laser amplitude noise. Two-tone tests have been used to demonstrate an intermodulation-free dynamic range of 90 dB. We have also used optical sampling to directly downsample frequency-modulated chirp signals having 1 GHz bandwidth on an X-band (10 GHz) microwave carrier. The bandwidth of the technique is extended by optically distributing the post-sampling pulses to an array of time-interleaved electronic quantizers. Using high-extinction 1-to-8 LiNbO3 optical time-division demultiplexers to perform the optical distribution, we have demonstrated a 500 MS/s ADC having 10 effective bits of resolution and a spur-free dynamic range in excess of 70 dB.
Photon-Counting Receiver Using InP Avalanche Photodiodes
Simon Verghese, PhD
Physics
University of California–Berkeley
The Mars Laser Communications Demonstration (MLCD) was planned as a NASA-sponsored optical communications experiment from Mars to Earth. Although the satellite will no longer fly in 2009, the required technologies for both the space segment and the ground segment were developed at MIT Lincoln Laboratory and the Jet Propulsion Laboratory. One of the Earth-based receivers was to have four 0.8 m telescopes, each of which could collect light on its own 8 × 8 array of InP-based avalanche photodiodes (APDs). This talk describes the APDs and the associated read-out integrated circuit (IC) that can collect signal photons from Mars. The read-out IC was a full-custom design with shared heritage from LADAR focal plane arrays. It includes high-speed timing circuits as well as control circuits that operate autonomously behind each pixel.
CANARY B-Cell Sensor for Rapid, Sensitive Identification of Pathogens| James D. Harper, PhD Biochemistry Massachusetts Institute of Technology |
Martha S. Petrovick, PhD Cell & Developmental Biology Harvard Medical School |
| Frances E. Nargi, PhD Pathobiology University of Connecticut |
Eric D. Schwoebel, PhD Molecular & Cell Biology Baylor College of Medicine |
Todd H. Rider, PhD
Electrical Engineering
Massachusetts Institute of Technology
A novel type of biosensor for rapid pathogen identification is being developed that uses B cells, or white blood cells, as the sensing elements. B cells, which function within the body to detect pathogens, respond in a fashion that is much more rapid, sensitive, and specific than currently available man-made sensors. Our sensor cells are genetically engineered to selectively bind to specific pathogens and then emit photons starting in less than a second after the pathogen is bound. A biochemical signal-amplification mechanism inside the cells provides an enhanced response that enables detection and identification of even a few pathogenic particles. The rapid identification of pathogens that this technology enables should be particularly important for biological-warfare agent detection, medical diagnostics, food safety assurance, environmental monitoring, and other applications.
Quantum Computation and Atomic Physics with Persistent-Current Qubits
William D. Oliver, PhD
Electrical Engineering
Stanford University
Superconducting persistent-current qubits are tunable artificial atoms with multiple energy levels, and they are a strong candidate technology for quantum computing applications. This talk will present recent experimental demonstrations of single-photon and multiphoton dynamics, including an analog to Mach-Zehnder interferometry with MIT Lincoln Laboratory’s qubit state. In this experiment, the qubit’s ground and first-excited states exhibit an avoided crossing. Strongly driving the qubit with harmonic excitation sweeps it through the avoided crossing two times per period. The induced Landau-Zener transitions act as coherent beamsplitters, and the accumulated phase between transitions varies with microwave amplitude. Quantum interference fringes for 1…50 photon excitations have been observed. This talk will present and discuss these experimental results.
Three-Dimensional Imaging Using Avalanche Photodiode Arrays
Brian F. Aull, PhD
Electrical Engineering
Massachusetts Institute of Technology
This presentation discusses the development of arrays of silicon avalanche photodiodes integrated with CMOS timing circuits and the application of these arrays to systems that capture three-dimensional images using laser radar techniques. The avalanche photodiodes are operated in Geiger mode; they are biased above the avalanche breakdown voltage so that the detection of a single photon leads to a discharge that can directly trigger a digital circuit. The CMOS circuits to which the photodiodes are connected contain high-speed counters that measure the times when the detection events occur. Examples are presented of three-dimensional images from a laser radar system that uses a Geiger-mode avalanche photodiode array.
Novel Detector Technology for Challenging Time-Dependent Imaging Applications
Bernard B. Kosicki, PhD
Physics
Harvard University
For most imaging applications, movement in the image is a concern after some exposure time threshold. The Advanced Imaging Group at MIT Lincoln Laboratory has developed three novel imaging devices to handle image motion in different ways. The orthogonal-transfer charge-coupled device (CCD) is designed to image on scenes with continuous translational motions, such as those caused by camera vibration or first-order atmospheric distortions. This device allows exposure times much longer than those allowed by a typical CCD imager, which, to avoid blur, must allow short exposure times. An electronic shutter has been developed for scientific CCDs. MIT Lincoln Laboratory has used this shutter to create an imager that captures, and stores in the pixel, several consecutive frames at 500 ns frame intervals while still having high quantum efficiency and low read noise. As a final example, an imager has been developed to record time-of-arrival of the first photon to each pixel with time resolution of 250 ps. This imager utilizes a single-photon-sensitive Geiger-mode avalanche-photodiode (APD) array as the light sensing layer. Each APD is connected to a high-speed digital timing circuit located directly below it. This imager has enabled construction of a flash LADAR system capable of creating three-dimensional images with 5 cm depth resolution.
High-Speed Solid-State Imager Technology
Dennis D. Rathman, PhD
Physics
Lehigh University
Electronically shuttered solid-state imagers are being developed for high-speed imaging applications. A 5 × 5 cm, 512 × 512 pixel, multiframe charge-coupled device (CCD) imager has been fabricated that collects four sequential image frames at megahertz rates. To operate at fast frame rates with high sensitivity, the imager uses an electronic shutter technology designed for back-illuminated CCDs. The design concept and test results are described for the burst-frame-rate imager.
CMOS-based solid-state imager technology has the promise for creating large-format X-ray detectors with short exposure times (100 ps to 1 ns). For example, CMOS transistors have switching speeds of tens of picoseconds needed for high-speed sampling circuits. A 64 × 64 pixel test circuit has been designed and fabricated in 0.18 µm CMOS technology to investigate high-speed imaging for large-format detectors. Several features have been integrated into the circuit architecture to achieve fast signal propagation with low skew and jitter for simultaneous pixel exposure times. These features include an H tree clock distribution, single-edge trigger propagation with local and global repeaters, local exposure control, and current-switching sampling circuits. Also, a unique photodiode structure that has a fast response of less than 100 ps has been designed for bump-bond integration with the CMOS readout. The photodetector has features intended to maintain the fast performance up to peak X-ray fluences of 1017 photons/sec•cm2. The CMOS readout and photodiode design philosophy will be discussed.
Three-Dimensional Integration Technology for Advanced Focal Planes and Integrated Circuits
| Brian F. Aull, PhD Electrical Engineering Massachusetts Institute of Technology |
James A. Burns, PhD Physics University of Vermont |
| Chenson K. Chen, PhD Physics University of California, Berkeley |
Craig L. Keast, PhD Electrical Engineering Massachusetts Institute of Technology |
| Jeffrey M. Knecht, MS Physics Northeastern University |
Vyshnavi Suntharalingam, PhD Engineering Science Pennsylvania State University |
| Brian M. Tyrrell, MS Electrical Engineering and Computer Science Massachusetts Institute of Technology |
Bruce D. Wheeler, BS Mechanical Engineering University of Massachusetts, Lowell |
| Peter W. Wyatt Engineering and Applied Sciences Yale Univeristy |
Dana-Ruth W. Yost, BS Materials Engineering Cornell University |
Keith Warner |
|
Over the last five years, MIT Lincoln Laboratory has developed a threedimensional circuit integration technology that exploits the advantages of silicon-on-insulator technology to enable wafer-level stacking and micrometer-scale electrical interconnection of fully fabricated circuit wafers [1].
Advanced focal-plane arrays have been the first applications to exploit the benefits of this three-dimensional integration technology because the massively parallel information flow present in two-dimensional imaging arrays maps very nicely into a three-dimensional computational structure as information flows from circuit tier to circuit tier in the z-direction. To date, the Laboratory’s three-dimensional integration technology has been used to fabricate four different focal planes, including a two-tier 64 × 64 imager with fully parallel per-pixel A/D conversion [2]; a three-tier 640 × 480 imager consisting of an imaging tier, an A/D conversion tier, and a digital signal processing tier; two-tier 1024 × 1024 pixel, four-side-abuttable imaging modules for tiling large mosaic focal planes [3]; and a three-tier Geigermode avalanche photodiode (APD) three-dimensional LIDAR array, using a 30-volt avalanche-photodiode tier, a 3.3-volt CMOS tier, and a 1.5-volt CMOS tier [4].
Recently, the three-dimensional integration technology has been made available to the circuit-design research community through Multiproject fabrication runs sponsored by the Defense Advanced Research Projects Agency. The first Multiproject Run (3DL1) completed fabrication in early 2006 and included over 30 different circuit designs from 21 different research groups. Three-dimensional circuit concepts explored in this run included stacked memories, field programmable gate arrays, and mixed-signal circuits. The second Multiproject Run (3DM2) is currently in fabrication.
This talk will provide a brief overview of MIT Lincoln Laboratory’s three-dimensional integration process, discuss some of the focal-plane applications in which the technology is being applied, and provide a summary of some of the Multiproject Run circuit results.[1] J.A. Burns, et al., “A Wafer-Scale 3-D Circuit Integration Technology,” IEEE Transactions on Electron Devices, Vol. 53, No. 10, pp. 2507–2516, October 2006.
[2] J.A. Burns, et al., “Three-dimensional Integrated Circuits for Low Power, High Bandwidth Systems on a Chip,” 2001 ISSCC International Solid-State Circuits Conference, Digest of Technical Papers, Vol. 44, pp. 268–269, February 2001.
[3] V. Suntharalingam, et al., “Megapixel CMOS Image Sensor Fabricated in Three-Dimensional Integrated Circuit Technology,” 2005 ISSCC International Solid-State Circuits Conference, Digest of Technical Papers, Vol. 48, pp. 356– 357, February 2005.
[4] B. Aull, et al., “Laser Radar Imager Based on 3D Integration of Geiger-Mode Avalanche Photodiodes with Two SOI Timing Circuit Layers,” 2006 ISSCC International Solid-State Circuits Conference, Digest of Technical Papers, Vol. 49, pp. 304–305, February 2006.
MIMO Wireless Communication
Daniel W. Bliss, PhD
Physics
University of California–San Diego
Wireless communication is playing a role of ever-increasing importance in our everyday lives. The need for higher data rates within the confines of limited spectrum allocation has motivated the investigation of high-spectral-efficiency communication. These communication links are often in non-line-of-sight complicated multipath environments.
Wireless communication using multiple-input multiple-output (MIMO) systems enables increased achievable spectral efficiency and reliability for a given total transmit power. The increased capacity is achieved through the introduction of antenna arrays at both transmitter and receiver. These arrays are used to take advantage of the multiple spatial modes provided by the complicated multipath environment. Space-time coding describes the waveforms employed by MIMO systems to approach theoretical capacity.
In this talk, an introduction to MIMO communication is provided. Sensitivity of theoretical capacity to environmental variations is considered. These environmental factors include channel complexity, channel estimation errors, and external interference. Channel phenomenology and its effect on capacity is investigated, using both physical models and experimental data. Parametric techniques used to model the experimental results are also introduced. Performance results for space-time turbo coding techniques are presented as a function of channel characteristics and receiver design.
Waveform Design for Airborne Networks
Frederick J. Block, PhD
Electrical Engineering
Clemson University
Airborne networks are expected to play a significant role in future military communications. Successful deployment of these ad hoc networks requires overcoming many unique challenges. For example, nodes are often separated by great distances and can be highly mobile. In addition to multiple-access interference from other radios in the network, interference from jammers located over a wide geographic region may be able to reach a receiver because of its high altitude. The presentation gives an overview of channel models for airborne networks and examines the trade-offs that must be made when choosing the modulation, coding, and channel access protocol.
Automated Topology Control for Wideband Directional Links in Airborne Military Networks
Wayne M. Bynoe, PhD
Computer Engineering
Boston University
Future airborne and military operations will rely on merging wideband directional RF and optical links for machine-to-machine networking. Management and control for such directional wireless links on mobile platforms is a significant challenge that does not arise in fixed terrestrial networks. As a consequence, new protocols and approaches are needed. In this talk, an architectural framework for control of wideband directional links in airborne military networks is presented. Three distributed algorithms for automated topology management that have been devised and prototyped at MIT Lincoln Laboratory are compared. Results and experience from simulations, emulations, and field tests are presented.
New Approaches to Automatic Speaker Recognition| Joseph P. Campbell, PhD Electrical Engineering Oklahoma State University |
Pedro A. Torres-Carrasquillo, PhD Electrical Engineering Michigan State University |
The area of automatic speaker recognition has been dominated by systems using only short-term, low-level acoustic information, such as cepstral features. While these systems have produced reasonably low error rates, they ignore other levels of information beyond low-level acoustics that convey speaker information. This seminar will include a tutorial and focus on late-breaking research that exploits high-level information, e.g., idiosyncratic word usage and pronunciation, in automatic speaker recognition systems to add robustness and improve accuracy.
Implementation Considerations for Wideband Wireless Communications
Nancy B. List, PhD
Electrical Engineering
Georgia Institute of Technology
Unexpected technical challenges often arise in the process of transferring technology from theory into practical applications. It is well known that modulator distortion causes problems for the transmission of communications signals. Less obvious, however, is the effect of modulator distortion on signals used for time tracking in wireless systems requiring strict timing control. Narrowband tracking signals are often used to synchronize systems transmitting wideband communications signals. While narrowband tracking signals may be less sensitive than communications signals to many types of distortion, they are particularly sensitive to group delay variation. As a result, relatively small levels of group delay variation across the frequency band can cause unexpected overall system degradation. This presentation will describe the real-world challenges of time-tracking in frequency-hopped satellite communications systems transmitting signals at high data rates, as well as practical methods to analyze and overcome these challenges.
Laser Communications Transceiver Design
David O. Caplan, PhD
Electrical Engineering
Northwestern University
Optical communications has provided unprecedented capacity in modern networks and has fueled the rapid growth of the Internet. Until recently, the sensitivity and efficiency of optical transmitters and receivers have not been driving factors in the buildup of fiber-optical networks. But as the demand for bandwidth approaches the limitations of current communication systems, more sensitive receivers can provide a means for improving optical network performance in terms of both power and bandwidth efficiency. High-sensitivity receivers can also reduce mid-span amplifier requirements, can extend link distances, and are especially beneficial for free-space optical communications since improvements in receiver sensitivity directly reduce transmitted power requirements.
This talk will provide an overview of optical transmitter and receiver designs, from relatively simple direct detection systems used in short terrestrial fiber-optic links to sophisticated near quantum-limited systems suitable for deep-space-based optical links. The basic characteristics of optical sources, modulators, amplifiers, detectors, and associated noise sources will be discussed along with some of the unique properties that distinguish optical communication systems and components from their RF counterparts. Also presented will be practical trade-offs and implementation issues that arise from using various technologies, modulation formats, and coding in both free-space and fiber-optic links.
Superconducting Nanowire Detectors for Photon-Counting Optical Communications at Gigabit/second Data Rates
Andrew J. Kerman, PhD
Physics
Stanford University
Optical receivers capable of detecting single photons can be used to provide a dramatic increase in the efficiency of optical communications over conventional methods, opening the possibility of communication over extremely large distances. Already underway at MIT Lincoln Laboratory is a demonstration of photon-counting optical communications between a satellite in Mars orbit and an Earth-based receiver; even larger interplanetary distances can be envisioned. So far, the most sensitive single-photon detectors in the near infrared region used for communications are InGaAs Geiger-mode avalanche photodiodes; however, these devices suffer from several important limitations, including a long reset time after a detection event, high dark count rates, and large timing jitter, the combination of which limits achievable data rates to ~tens of megabits/second.
In this presentation, a new optical detector technology based on nanowires lithographically patterned onto ultrathin, superconducting NbN films will be discussed. When biased with a DC current close to the critical current above which the material becomes resistive, these nanowires become highly sensitive to incident photons and exhibit sufficiently fast reset times, low dark count rates, and low timing jitter to support photon-counting optical communications at gigabit/s data rates. Their high speed and timing resolution may also be of interest for experiments in solid-state and biological physics. We will discuss the physics of these devices, their fabrication, characterization, and integration into an optical receiver architecture, as well as future directions and potential improvements.
Reusable Communications Waveform Development
Wayne G. Phoel, PhD
Electrical Engineering
Northwestern University
The concept of a software-defined radio is powerful: use a single radio device to host any of several sets of radio protocols and signal processing simply through different loads of software. The challenges are in designing the device to be flexible enough to satisfy the range of requirements the different waveforms have and in writing the software in such a way that it can be ported to different hardware platforms while maintaining interoperability. The problems become even more challenging when the waveform complexity increases to require firmware implementations. This seminar discusses recent work on the design of a high-performance, reconfigurable hardware architecture and development of portable firmware for advanced communications waveforms. Strategies for ensuring interoperability among different hardware platforms hosting the same waveform software and firmware will also be discussed.
QoS and Cross-Layer Optimization for Satellite Communications Networks
Jeffrey S. Wysocarski, PhD
Electrical Engineering
Clemson University
To efficiently utilize limited RF resources, future packet-switched satellite networks will dynamically allocate resources on the uplink and downlink. Designing the resource-allocation algorithms to maximize link-layer efficiency is insufficient. The resource-allocation algorithms must work cooperatively with the network layer and transport layer to optimize network layer performance and provide quality of service (QoS) to applications and users. Several mechanisms for facilitating this required cooperation between the layers are presented. The individual roles and actions of the layers as well as their interaction are defined. Router QoS schedulers that continue to provide service differentiation in the presence of link variations are illustrated, and downlink scheduling architectures that provide terminal QoS guarantees are demonstrated. Finally, the interaction between TCP and the dynamic resource-allocation algorithms is investigated, leading to suggested modifications of either the resource-allocation algorithms, the TCP protocol, or both.
Dynamic Link Adaptation for Satellite Communications
Huan Yao, PhD
Electrical Engineering
Massachusetts Institute of Technology
Future protected military satellite communications will continue to use high transmission frequencies to capitalize on the large amounts of available bandwidth. However, the data flowing through these satellites will transition from the circuit-switched traffic of today’s satellite systems to Internet-like packet traffic. One of the main differences in migrating to packet-switched communications is that the traffic will become bursty (i.e., the data rate from particular users will not be constant). The variation in data rate is only one of the potential system variations. At the frequencies of interest, rain and other weather phenomena can introduce significant path attenuation for relatively short time periods. However, the current systems are designed to always provide sufficient margin even when it is not raining. The focus of this seminar is the design of a future satellite system that autonomously reacts to changes in link conditions and offered traffic. This automatic adaptation drastically improves the overall system capacity and the service that can be provided to ground terminals.
Managing Large-Scale Information Operations Tests
Tamara H. Yu, MEng
Computer Science
Massachusetts Institute of Technology
The Lincoln Adaptable Real-time Information Assurance Testbed (LARIAT) allows researchers and government buyers to test the effectiveness of information operations (IO) technologies in a realistic, closed network environment. As IO tests scale up in size and complexity, it is increasingly difficult to set up and validate testbed networks and to monitor test progress. Recent efforts to improve the LARIAT software suite focus on the ease of use of the software, especially in large tests, by introducing a set of user interface and visualization tools that help users manage testbed resources, rapidly install and configure software, and control and monitor the testbed from a central location.
Discovering Near-Earth Asteroids at MIT Lincoln Laboratory
J. Scott Stuart, PhD
Earth, Atmospheric, & Planetary Sciences
Massachusetts Institute of Technology
The Lincoln Near-Earth Asteroid Research (LINEAR) program has discovered more than half of all near-Earth asteroids (NEAs) ever discovered. Since beginning full-time operations in 1998, it has provided 60% of the worldwide discoveries of new NEAs from its site near Socorro, New Mexico. On top of its success in discovering NEAs, LINEAR has become the leading ground-based discoverer of comets, with more than one hundred and fifty comets now named “LINEAR.” LINEAR discovers many comets when they are far away from the Sun on their inbound trajectory, thus allowing observation of the heating process missed when comets are discovered closer to the Sun. LINEAR now operates two wide-area search telescopes with 1 m apertures and has recently added a 31 in aperture telescope dedicated to automated follow-up of possible NEAs detected by the search telescopes. The LINEAR program originated as a new application of technology developed by MIT Lincoln Laboratory to provide the United States Air Force with enhanced capability to track spacecraft. This successful technology migration has resulted in an improved understanding of the NEAs with LINEAR data providing the basis for the best analyses of the asteroid impact risk to the Earth. This talk provides an overview of the LINEAR program, including recent enhancements to the LINEAR system, the productivity of the program, the scientific results gleaned from LINEAR data, and descriptions of some of the more interesting objects discovered.
Remote Real-Time Radar Imaging
Jennifer L. Sulyma, MS
Mathematics
University of Massachusetts–Amherst
MIT Lincoln Laboratory has recently completed a development effort known as the Wideband Networked Sensors (WNS) project. The goal of this DARPA-funded project was to demonstrate a communication-intensive military application—remote real-time radar imaging of complex satellites and sensor fusion—using the BoSSNET test bed, an all-optical, wide-area network that connects Boston and Washington, D.C.
The X-band Haystack Long Range Imaging Radar (LRIR) and Ku-band Haystack Auxiliary Radar (HAX) located in Tyngsboro, Massachusetts, were connected via a 60-mile-long fiber network to a remote processing center located at MIT Lincoln Laboratory in Lexington, Massachusetts. Unprocessed radar data (I&Q) were transmitted through the network to the center, where data from the two bands were fused and ultra-wideband images were generated in real time and displayed. The images were sent to the BoSSNET and looped back to simulate its capability to support wideband communication long-distance link.
The remote real-time imaging system consisted of three separate processing pipelines: one image stream per radar and one for the virtual ultra-wideband radar to demonstrate the sparse band fusion processing. The complex processing of the individual radars data was accomplished in real time, while the fused ultra-wideband images were processed in real time and near-real time.
This seminar will discuss the development effort, system configuration, and processing software and highlight some recent remote real-time imaging demonstrations.
A System for Predicting Close Approaches and Potential Collisions in Geosynchronous Orbits| Richard I. Abbot, PhD Physics/Meteorology University of Texas–Austin |
Miquela C. Vigil, PhD Earth, Atmospheric, & Planetary Sciences Massachusetts Institute of Technology |
The geosynchronous orbit is getting crowded with over 300 active, revenue-producing large satellites and over 500 inactive resident space objects that pose a physical collision threat to the active satellites. The in situ demise of a particular satellite, Telstar 401, followed by a similar demise of SOLIDARIDAD 1, initiated a research and development effort at MIT Lincoln Laboratory to address this threat. This work with commercial satellite operators is accomplished using the mechanism of Cooperative Research and Development Agreements. Initial work to detect and warn of close approaches with these two failed satellites led to more extensive research on the collision threat over the entire geosynchronous belt. It is apparent that
- There is a significant probability of collision;
- The probability has increased considerably in the last decade or so;
- The continuing failure of geosynchronous satellites and injection of rocket bodies into or near geosynchronous orbit will increase the threat; and
- Debris in or near geosynchronous orbit poses another problem that has to be addressed.
This seminar surveys what has been achieved so far in predicting the threat and protecting satellites. An assessment of the probability of collision is presented as well as a description of the Geosynchronous Monitoring and Warning System (GMWS). The operations of the GMWS, as well as some of the results achieved so far, are described. Areas of current research are mentioned.
Environmental Applications of Hyperspectral Data| Hsiao-hua K. Burke, PhD Physics Rice University |
Michael K. Griffin, PhD Meteorology University of Utah |
| Joseph W. Snow, PhD Atmospheric Sciences University of Virginia |
Carolyn A. Upham, MS Atmospheric Sciences University of Massachusetts–Lowell |
In this presentation, the application of visible and IR spectral information to atmospheric characterization as well as surface feature retrieval will be discussed and illustrated with recent data.
AVIRIS and Hyperion data are utilized. The Airborne Visible-InfraRed Imaging Spectrometer (AVIRIS) sensor contains 224 bands, each with a spectral bandwidth of approximately 10 nm, allowing it to cover the entire range between 0.4 and 2.5 µm. For a NASA ER-2 flight altitude of 20 km, each pixel is 20 m in size, yielding a ground swath width of approximately 10 km. The EO-1 satellite is part of NASA’s New Millennium Program (NMP). Hyperion, one of the sensors of the EO-1 payload, is a hyperspectral imager with AVIRIS-like spectral coverage and resolution and 30 m ground pixel resolution. The instrument images a 7.5 km by 100 km area per image. Hyperion has been the only space-borne HSI data source since the launch of EO-1 in late 2000.
Examples of data applications include atmospheric sounding, cloud depiction, aerosol characterization, water vapor retrieval, and surface feature delineation. It is illustrated that though each application may only require a few spectral bands, the ultimate strength of HSI exploitation lies in the simultaneous and adaptive retrievals of atmospheric and surface features. Interrelationships among different bands are also demonstrated, and these are the physical basis for the optimal exploitation of spectral information.
Retrieval of Atmospheric Temperature and Moisture Profiles from Hyperspectral Sounding Data Using a Projected Principal Components Transform and a Neural Network
William J. Blackwell, PhD
Electrical Engineering
Massachusetts Institute of Technology
A novel statistical method for the retrieval of atmospheric temperature and moisture (relative humidity) profiles has been developed and evaluated with simulated clear-air hyperspectral sounding data. The accuracies of the estimates produced by the algorithm meet or exceed (in some cases by a factor of two) the accuracies of the estimates from traditional iterated minimum variance retrieval techniques while requiring less computation. The algorithm is implemented in two stages. First, a projected principal components (PPC) transform is used to reduce the dimensionality of and optimally extract geophysical profile information from the spectral radiance data. Second, an artificial feedforward neural network (NN) is used to estimate the desired geophysical parameters from the projected principal components.
The performance of this method (henceforth referred to as the PPC/NN method) was evaluated using simulated clear-air observations from the 2378-channel Atmospheric InfraRed Sounder. Separate training and validation profile data were selected from the NOAA88b radiosonde set of approximately 7500 profiles. Surface, solar, and instrument effects were modeled.
It was found that the PPC/NN method has a number of advantages over traditional statistical and physical/iterative hyperspectral profile retrieval techniques. Neural-network estimates based on the PPC transform were significantly more accurate than neural-network estimates obtained with conventional principal components techniques, including the Karhunen-Loeve transform and the noise-adjusted principal components transform. The retrieval accuracy of PPC/NN was also superior to that of a principal components regression technique.
One particularly noteworthy result of the present work is a comparative study between the PPC/NN method and an iterated minimum-variance (IMV) method. The temperature profile retrieval accuracy of both methods is similar, but the relative humidity profile retrieval accuracy of PPC/NN was greater than that of the IMV method at all altitudes and substantially better near the surface.
Space Surveillance with the Space-Based Visible Sensor
Jayant Sharma, PhD
Aerospace Engineering
University of Texas–Austin
The Midcourse Space Experiment satellite was launched in 1996. A principal sensor on board the satellite is the Space-Based Visible (SBV) sensor, a visible-band, electro-optical camera designed at MIT Lincoln Laboratory to perform the first technical and functional demonstration of space-based space surveillance. The principal task of the SBV sensor is to gather tracking data on satellites. In 1997, after the successful technology-demonstration phase of the mission, the SBV sensor was transitioned from an experimental sensor to a dedicated sensor in the Space Surveillance Network. The Space-Based Space Surveillance Operations is now providing the Space Surveillance Network with the first operational space-based space surveillance sensor. With its orbital location, wide field of view, and high metric accuracy, the SBV sensor has made significant contributions to the Space Surveillance Network. This talk will introduce the topic of satellite tracking and the development and evolution of satellite tracking from a space-based sensor. The performance of the SBV sensor and the application of its observations to satellite tracking will be demonstrated.
Bayesian Inference Approach to Learning Coordinated Traffic Behavior for Non-tracking Sensors
Lawrence A. Bush, MS
Computer Science
Rensselaer Polytechnic Institute
MIT Lincoln Laboratory has developed a method for learning and detecting coordinated traffic behavior for situations in which tracking is infeasible. For example, tracking in large areas with dense traffic is very demanding of sensor resources, and maintaining track of interacting objects is extremely difficult. Therefore, non-tracking methods for interpreting intentional traffic coordination have been explored.
The approach is to statistically model different traffic behavior classes, at a particular location, in order to detect the activity of interest and then combine the results from multiple locations using a reasoning structure. The seminar will present a method for learning this reasoning structure from the data using a Bayesian Network structure search, then using the Bayesian Network to infer the overall situation across multiple sites. This approach emphasizes evidence accumulation and continuous learning, which lends strong support to the proposed multi-input computational framework.
MIT Lincoln Laboratory’s research is applied to wide-area persistent surveillance using moving target indication (MTI) radar data. MTI is a radar data processing technique for detecting moving vehicles. The MIT Lincoln Laboratory approach was tested by collecting MTI data while running multiple experimental military ground scenarios, each involving coordinated activity over multiple sites. By using this approach, the overall behavior classes were reliably identified.
This talk will include a tutorial of the related statistical modeling process and Bayesian inference technique. This will be followed by an evaluation of the Laboratory’s results, which demonstrate the ability of this technique to identify military activity.
Tracking in a Maritime Context
Leslie Servi, PhD
Electrical Engineering
Harvard University
As noted in the 2006 Quadrennial Defense Review, there is a national consensus to improve our maritime domain awareness to impede terrorism, piracy, or illegal drug activities. Significant problems in tracking vessels of interest on the high seas are limited sensors, long lapses in position updates, and imprecise feature measurements. Advanced technologies can help improve traditional tracking performance by focusing on the following characteristics: (1) nonlinear dynamics, (2) non-Gaussian measurements of location and/or features, (3) associating imperfect feature measurements with vessels, (4) discrete multiple modes of dynamics (fishing vs. cruising), and (5) features that may evolve in time.
A tracking algorithm that handles such characteristics will be presented along with a discussion of methods to validate the system performance.
Feature Extraction for Classification: Class-Independent Statistics vs. Class-Dependent Statistics
| John L. Weatherwax, PhD Applied Mathematics Massachusetts Institute of Technology |
Virginia K. Hafer, BA Physics Wellesley College |
There currently exist a large number of algorithms aimed at reducing the dimensionality of a feature set used for classification by relying on the statistical properties of the underlying distribution of data. Three common and well-known examples of this type of technique are Principal Component Analysis (PCA); Multiple Discriminant Analysis (MDA), also known as Linear Discriminant Analysis (LDA); and Singular Value Decomposition (SVD). Each technique has its own operating conditions and a priori assumptions about the statistical nature of the data. In particular, for classification of the techniques mentioned, only the MDA algorithm explicitly uses class information. Recently, there has been interest in comparing PCA against MDA. It is generally felt that the projected feature subset produced by MDA is better for classification purposes since the MDA algorithm finds a linear transformation that maximizes a measure of class separation. This talk will empirically compare and contrast the performance of the PCA, MDA, and SVD techniques on a variety of standard datasets. In addition, two additional algorithms that incorporate class-specific information into their constructions will be introduced: Classwise Principal Component Analysis (C-PCA) and Classwise Singular Value Decomposition (C-SVD), based on PCA and SVD, respectively. It will be demonstrated empirically that these new algorithms classify data better than do their standard PCA and SVD counterparts. In addition to classification, each of the algorithms presented has a natural method for dimensionality reduction via eigenspace truncation. Also presented will be results examining how classification performance degrades as the number of implicit algorithm-specific degrees of freedom are decreased. Results demonstrate that the C-PCA and C-SVD techniques allow one to potentially do significant dimensionality reduction with negligible impact in classification accuracy.