Tufts University  |  School of Engineering  |  Find People  |  Give  | 


Archives: Spring 2014

Smarter Sensing for Critical Infrastructures: The Intersection of Physical Principles, Sparse Models, and Statistical Signal Processing

Acoustic signal processing methods are important tools for solving problems in civil engineering, aerospace engineering, medicine, oceanography, and seismology. In these fields, acoustics signals are used to monitor bridges and airplanes, noninvasively diagnose ailments and illnesses, track objects and animals from great distances, and predict oncoming seismic events.

For civil and aerospace engineering applications, ultrasonic guided waves (waves that are “guided” by the geometry of the environment) are transmitted through critical infrastructures, such as bridges, pipelines, and airplanes, to test materials and structures for damage and degradation. This approach is used in prevent catastrophic failures in important infrastructures, including transportation systems, power plants, and resource distribution pipeline networks. To detect, locate, and evaluate damage in these structures, acoustic guided waves are measured and analyzed using various signal processing strategies. However, successfully detecting and locating damage is challenging because the waves propagate with complex, dispersive behavior that distorts the shape and phase of waves as they travel through the medium.

I present a smart infrastructure framework that overcomes these challenges by representing the general solution to the wave equation as a sparse model and recovering that model with compressive sensing tools to learn the complex characteristics of the waves. With experimental data, I demonstrate how to find these characteristics and then how to utilize the properties to improve our analysis of complex guided waves and our evaluation of critical infrastructures with statistical signal processing methods. Results show significant improvements in detection and localization performance over conventional approaches.

Joel B. Harley received the B.S. degree in electrical engineering from Tufts University, Medford, MA, in 2008 and a M.S. degree in electrical and computer engineering from Carnegie Mellon University, Pittsburgh, PA in 2011. He is currently working toward a Ph.D. degree in electrical and computer engineering at Carnegie Mellon University, Pittsburgh, PA. His interests include the integration of complex wave propagation models with novel signal processing, machine learning, and big data methods for applications in cyber-physical systems, structural health monitoring, nondestructive evaluation, and other fields.

Mr. Harley is a recipient of the 2009 National Defense Science and Engineering Graduate (NDSEG) Fellowship, the 2009 National Science Foundation (NSF) Graduate Research Fellowship, the 2009 Department of Homeland Security Graduate Fellowship (declined), and the 2008 Lamme/Westinghouse Electrical and Computer Engineering Graduate Fellowship. He has published more than 30 technical journal and conference papers, including four best student papers. He is a student representative for the IEEE Ultrasonics, Ferroelectrics, and Frequency Control Society and a member of the IEEE Signal Processing Society and the Acoustical Society of America.

The Design of a Radically Different Solar Inverter by a Mechanical Engineer or Certain Advantages of Innovation from Left Field

Almost all solar photovoltaic inverters in use today are based on a pulse width modulation topology. These inverters also represent a major cost component in a photovoltaic system, and their percentage of total system cost has been increasing as solar module prices have come down in recent years. There also has been considerable activity in the industry to employ distributed electronics to optimize and control the performance of individual modules in a string, something that cannot be done by central string inverters.

Recognizing these issues, an effort was made to design a solar inverter that would be much less expensive and would incorporate the desired local control features. This task was undertaken by a mechanical engineer with only rudimentary knowledge of electronics.

The task was successful: the new design, proven out in testing over three generations of prototypes, turned out to be highly efficient, potentially low-cost and highly reliable. Moreover, the technology is able to precisely control the maximum power point of individual modules, provide fire, installation and arcing safety, bypass nonperforming modules and provide real-time performance feedback from every module in the string.

However, the important take-away from this task is not the technology itself, although it is sort of fun. Rather, it is instructive on the benefits of not having a preconceived notion of how things should be done.

Privacy-Security Tradeoffs in Biometric Template Protection

Biometric Template Protection refers to a class of authentication or identity verification methods that prevent biometrics from being compromised when biometric systems are attacked. This is an exciting area of research and myriad template protection proposals have appeared in the last few years. The current state of the art owes its existence to various fields: signal processing, cryptography, information theory and machine learning. Each of these fields has provided a different language of describing key concepts, different implementation methods, and different characterizations of performance, security and privacy. In this talk, we will describe a unified analysis framework for biometric template protection - a single abstract construction consisting of an encoding module and a decision module that allows us to give precise, implementation-agnostic definitions for measures of performance (accuracy and robustness), security and privacy.

For concreteness, we shall analyze one flavor of template protection, called a secure sketch (or a fuzzy sketch). At its most basic, this involves applying error correction coding (ECC) to an enrollment biometric, and storing only the parity symbols (not the enrollment biometric) in the biometric database. Thus, if the database is compromised, the adversary only gains access to the parity symbols. We will analyze the accuracy, robustness, security and privacy for both one-factor and two-factor versions of the secure sketch. We will use mutual information to quantify the information leaked about a user's biometric, when the biometric template and any other keys are compromised. Further, we will also quantify the probability of successfully attacking the system based on the compromised information. We will close by drawing attention to an important unsolved problem, viz., the design of template protection systems for practical scenarios in which a user has enrolled her biometric at multiple access control systems. In this case, compromising one access control system may increase the attacker's chances of discovering the user's biometric, or of compromising the other systems. This problem requires joint design of the ECC matrices used by each template protection system. A toy example reveals that, while the accuracy vs. robustness tradeoff of biometric systems is generally well understood, the security vs. privacy tradeoff turns out to have interesting and non-obvious properties.

Extension of Electron and Ion Microscopy to Four Dimensions

I will describe how advances in electron and ion microscopy have enabled extension from studies in two dimensions (i.e. the generation of two dimensional images) to four dimensions (i.e. generation of three dimensional images, and the evolution of microstructure as a function of time). In particular, I will show how kinetic processes in materials may be studied in real time and at nanoscale resolution through application of mechanical, thermal, electrical, optical and electrochemical in-situ in the transmission electron microscope. I will also discuss the development of three-dimensional focused ion beam techniques, and the potential extension to nano-scale chemical reconstructions by employing Ion-Induced Auger Electron Spectroscopy.

Rear Illumination Monolithically Integrated GaSb Photovoltaic Devices Grown on Semi-Insulating GaAs Substrate

The photovoltaic industry has been focusing on technology to develop PV cells for terrestrial applications. Back-contact structures have been widely studied as alternative solar cells for high efficiency. Particularly, a back-junction solar cell has been attractive since it reduces the grid shading loss and the low resistive loss in the metal grids. It includes metallization patterns with both p/n junctions on the rear side of the device and photon flux is received through the front surface. For metal contacts, a concept of the monolithic interconnected modules (MIMs) can be utilized in order to avoid large ohmic losses by achieving a high-voltage/low-current configuration with low series resistance. However, the complicated processing steps which include defining, isolating and contacting patterns on the single-sided surface have always been a challenge. For large area wafer scale MIM PV cells, it is necessary to develop reliable and accurate fabrication techniques in the area of metal interconnection. The work presented here explores the feasibility of fabricating MIM PV cells with single-sided metal contacts at the wafer scale using a device consisting of GaSb/GaAs bilayers grown on SI GaAs. In order to reduce electrical losses associated with series resistance, we processed MIM devices each consisting of 24 PV cells interconnected in series, resulting in voltage increase up to 6 volts.

Representation and Coding of Signal Distances: Comparing Signals in the Cloud

Signal representation theory and practice has focused on how to best represent a signal as efficiently as possible and minimize the distortion on the signal incurred by the representation. However, in many applications the processing stage only requires the extraction of specific information from the signal and the signal itself is not necessarily of interest. In such applications the representation should be information scalable, i.e., adaptable to efficiently represent only the information required by the processing.

In this talk, motivated by cloud-based image retrieval applications, we demonstrate that such information scalability can be achieved using appropriately designed signal embeddings. Combined with quantization, these embeddings are a perfect fit for inference applications with storage, processing or communication constraints, such as augmented reality. These embeddings capture all or part of the geometry of the signal space, as required for inference, at a very low bit-rate. Thanks to this property, we can reduce the storage or transmission rate by more than 50% when compared to existing approaches for image retrieval.

Quantum Nanophotonics and Nanomechanics with Diamond

Diamond possesses remarkable physical and chemical properties, and in many ways is the ultimate engineering material - "the engineer's best friend!" For example, it has high mechanical hardness and large Young's modulus, and is one of the best thermal conductors. Optically, diamond is transparent from the ultra-violet to infra-red, has a high refractive index (n = 2.4), strong optical nonlinearity and a wide variety of light-emitting defects. Finally, it is biocompatible and chemically inert, suitable for operation in harsh environment. These properties make diamond a highly desirable material for many applications, including high-frequency micro- and nano-electromechanical systems, nonlinear optics, magnetic and electric field sensing, biomedicine, and oil discovery. One particularly exciting application of diamond is in the field of quantum information science and technology, which promises realization of powerful quantum computers capable of tackling problems that cannot be solved using classical approaches, as well as realization of secure communication channels. At the heart of these applications are diamond's luminescent defects-color centers-and the nitrogen-vacancy (NV) color center in particular. This atomic system in the solid-state possesses all the essential elements for quantum technology, including storage, logic, and communication of quantum information.

I will review recent advances in nanotechnology that have enabled fabrication of nanoscale optical devices and chip-scale systems in diamond that can generate, manipulate, and store optical signals at the single-photon level. Examples include a room temperature source of single photons based on diamond nanowires1 (Figure A) and plasmonic appertures2, as well as single-photon generation and routing inside ring3 (Figure B) and photonic crystal resonators (Figure C) fabricated directly in diamond4. In addition to these quantum applications I will present our recent work on diamond based on-chip frequency combs, as well as diamond nanomechanical resonators (Figure D).

Analytical Models for Internet Content Distribution

Delivery of transactions over the Internet involves coordination between independent network actors. These networks actors have different economic incentives, precluding the Internet from evolving as a single system in a coordinated fashion. In this talk, we discuss the efficiencies that content distribution networks (CDNs) offer within this economic context. The Internet would not be able to support services at scale without these efficiencies.

The Akamai mapping control system directs how end-users transact with the Akamai content-delivery platform - matching web traffic to servers and doing it at an unprecedented scale. I will present analytical models that are the underpinnings of the real time control problems that Akamai solves to perform this function. These models operate in a hierarchical fashion, solving inter-related optimization problems at different time scales. I will describe how this decomposition allows Akamai to deliver the performance objectives expected by end-users while driving efficiencies in the way the Internet operates on a global level.

Flat Optics Based on Designer Metasurfaces

Metasurfaces based on sub-wavelength patterning have major potential for realizing arbitrary control of the wavefront of the diffracted light by achieving local control of the phase, amplitude and polarization. Our prototypes are based on optical antenna arrays which enable a new way to control the phase of the scattered light using phase discontinuities. We discuss novel devices based on this technique; a salient feature is the ability to create often with a single digital mask an arbitrary analog phase profile. A variety of flat optical components, including gradient metasurfaces, lenses, polarizers, vortex plates, coatings, holograms and couplers with polarization invariant coupling efficiency will be presented. Finally we show that thin film coatings with thickness much less that the wavelength exhibit strong interference effects associated with the finite optical losses. Such thin films have strong potential for novel detectors, solar cells as well as antireflection coatings and perfect absorbers. We conclude by discussing our perfect absorber work in vanadium oxide nanometer thick films on Sapphire, which behave as a natural tunable metamaterial near the metal insulator transition. Perfect thermal emission and large broadband negative differential thermal emittance have also been recently observed in these films.

Photovoltaic (PV) Technology

With a 1.45 eV band gap, a high absorption coefficient, and proven low-cost, high-volume manufacturing, CdTe-based photovoltaic (PV) technology accounted for more than 7% of the 100 GW of solar power generating capacity that was installed worldwide by the end of 2012. Copper doping of CdTe assists in forming back contacts to CdTe, but Cu is known to diffuse through the device and produce junction-shunting pathways. If long-term degradation of the photovoltaic performance could be eliminated without significant changes in the device architecture or manufacturing process flow, the levelized cost of electricity would be substantially reduced.

I will discuss results (1) that demonstrate that carbon single-wall nanotube (SWNT) films can make a high-performance electrical back contact to CdTe solar cells without the use of Cu. Back contacts formed with SWNT films showed improved open circuit voltage in comparison to cells fabricated with standard Cu/Au back contacts and, once overcoated with a thin metal layer, the solar-to-electric conversion efficiency was higher as well. The results are understood by considering that individual SWNTs within the film extend through the film's thickness and make barrier-free contacts to individual CdTe grains in the active layer of the device.

This work was supported by the U.S. Department of Energy under Award Number DE-SC0006349 and by faculty start-up funds from the University of Toledo, and was conducted in collaboration with the Wright Center for Photovoltaics Innovations and Commercialization.

Making the Mid-Infrared Nano with Designer Plasmonic Materials

The mid-infrared (mid-IR) spectral range (3-30µm) has become a burgeoning and dynamic field of research both for fundamental exploration as well as for more applied research in health and the environment, security and defense, communication, and sensing. At the same time, the areas of plasmonics and metamaterials have experienced explosive growth over the past decade, fueled in part by rapid developments in fabrication, characterization, computational science, and theory. Yet, the integration of plasmonic structures into mid-IR optical systems has been slower to evolve. While scaling metamaterial and plasmonic geometries to mid-IR wavelengths is actually fairly straightforward, replicating the near-IR and visible optical properties of constituent materials in plasmonic and metamaterial systems is less trivial, leading to very different behavior of scaled systems in these two wavelength ranges.

In this talk, I will discuss our group's recent work developing novel optoelectronic and plasmonic devices and structures for mid-IR applications. I will demonstrate the advantages and disadvantages of utilizing traditional plasmonic metals in mid-IR structures, and use this discussion to motivate our recent work with highly doped semiconductors as designer mid-IR metals for plasmonic, metamaterial, and epsilon-near-zero applications. In particular, I will focus on the promise of these new plasmonic materials for nano-scale confinement of micron-scale wavelengths, and for potential applications in sensing and thermal emissivity control. Recent results demonstrating all-semiconductor perfect absorbers and nano-antennas, as well as new types of mid-IR emitting quantum nanostructures, integrate-able with mid-IR plasmonic materials, will be presented.

Enhanced transmission through a subwavelength slit using epsilon-near-zero materials. Plots show experimental (top) and schematic (bottom) transmission for undoped GaAs (left) and highly-doped InAs (right) sub-layers with strong transmission enhancement seen at lENZ~8µm for TM polarized light transmitted through the doped InAs layer. Schematic of doped InAs nanopillar coated in weakly absorbing molecules. Background shows scanning electron micrograph of nanoantenna array (left) and finite element simulation of a single nanoantenna array at resonance (right). Inset shows reflection spectra for nanoantenna array before (dashed) and after (solid) coating with a thin (~50nm) layer of weakly absorbing polymer.

Ph.D. defense: Initiating and sustaining arrays of microplasmas: experiment and simulation

Non-thermal microplasmas have been intensively studied because of their ability to generate high electron density while maintaining low gas temperature, even at atmospheric pressure in the absence of any vacuum system. Microplasma arrays, powered by a single microwave source, can interact with large area substrate and are potentially useful for industrial large scale roll-to-roll coating. A challenge in microplasma array design, however, is that plasmas located far from the energy source receive a lower fraction of the available power and are reluctant to be ignited. This issue leads to limited dimensional scalability. This dissertation targets potential ways to facilitate microplasma ignition by 1) application of an external DC field 2) injection of seed electrons and 3) injection of photons shared from neighboring microplasmas. Experiments are specifically devised to isolate each individual contribution to plasma ignition. Results show that the addition of a DC electric field alone will rapidly sweep seed electrons to the plasma electrodes prior to avalanche breakdown and, counter-intuitively, causes higher ignition power. Electrons from adjacent microplasmas can be attracted by a positive DC potential and are shown to lower the ignition power, but only at low pressures. Photons from nearby microplasmas with energy larger than the work function of the electrode are shown to induce photoelectrons and substantially reduce the ignition power. A Monte-Carlo simulation is developed to simulate the ignition process. The simulation results are compared with experiments to confirm the ignition power reduction due to the addition of photoelectrons. These findings may be used to guide microplasma array designs that utilize electrons and photons from nearby plasmas within the array. These methods will help ignite plasmas near the edge of large arrays more readily, improve plasma array scalability and eventually lead to large microplasma arrays.

Smarter Naval Electrical Power Systems

The US Navy currently has two different classes of destroyer, the Arleigh Burke (DDG 51) class and the Zumwalt (DDG 1000) class. The Zumwalt class has an Integrated Power System (IPS) rated at 54 MW while at 20 knots. In 2008, the Zumwalt class was limited to three ships, and the Arleigh Burke class was restarted. The next generation Arleigh Burke class ships are expected to be built with 12 MW starting in 2016. With the Arleigh Burke class set to comprise most of the destroyer fleet in the US Navy's 30 year shipbuilding plan, there is a need to be more efficient and smarter about shipboard power use to enable new capabilities and reduce fuel consumption.

Sampling and Inference for Spatiotemporal Single-Photon Imaging

Resolving individual photons in space and time is the holy grail of optical imaging. Recent advances in materials, devices and fabrication technologies have led to an emerging class of solid-state sensors with single photon sensitivity. Thanks to their sub-nanosecond time resolution, and rapidly increasing spatial resolutions, these new single-photon sensors (SPS) have been a key enabling technology behind recent breakthroughs in several domains, including fluorescence-based bio-imaging, time-of-flight 3D computer vision, LIDAR, high-speed videography, and astronomy. In this talk, I will present models, theory, and algorithms in signal sampling and inference to address several challenges associated with the SPS. In particular, I will present our recent work on establishing the performance bounds of the SPS in acquiring light intensity fields; on time-sequential adaptive sensing schemes that allow one to push the imaging capabilities of SPS systems beyond the nominal limit imposed by current hardware; and on new image formation algorithms that can efficiently "decode" the massive bitstreams generated by the SPS.

Clutter Reduction for Geocollaborative Interfaces

As robotic capabilities become more autonomous and reliable, there is a desire to integrate robots into first response and military deployments. A benefit, but also a curse of robotic technology is the immense amount of information that is collected and relayed to the human operator via the interface. During a large, mass casualty event the overall response system incorporates information from multiple sources, including: first response personnel responding to the event, a priori information about the incident location, incident command, social networks, etc. Often, it is critical to provide access to all the information, even if it is not currently relevant to a particular human operator. Clutter emerges as a significant problem when all information is available and the human must attend to particular aspects. Most clutter reduction approaches require a priori classification of information, human classification in real time, or are random. None of these approaches are appropriate for high stress, unknown, and dynamic situations.

We have developed two methods for addressing different aspects of this problem. The General Visualization and Abstraction algorithm provides a means of automatically reducing clutter based on information type, novelty, relevance, spatial aspects and human operator role. Feature Sets is an approach that focuses on reducing clutter by considering the environmental context, geospatial relationships, temporal aspects, and semantic relationships. This presentation will focus on the problem challenges, the algorithms, associated results and future directions.

Dr. Julie A. Adams is an Associate Professor of Computer Science and Computer Engineering in the Electrical Engineering and Computer Science Department at Vanderbilt University. She is also the founder of the Human-Machine Teaming Laboratory. Prior to joining Vanderbilt, Dr. Adams was an Assistant Professor of Computer Science at Rochester Institute of Technology (RIT). Before returning to academia, she worked in Human Factors for Honeywell, Inc. and the Eastman Kodak Company from 1995 to 2000. Dr. Adams received her Ph.D. degree in Computer and Information Sciences in 1995 from the University of Pennsylvania (Penn), performing her research on human-robotic interaction for multi-robot systems in Penn's General Robotics, Automation, Sensing and Perception (GRASP) Laboratory. She received her M.S.E. degree in Computer and Information Sciences from the University of Pennsylvania, and her B.S. in Computer Science and B.B.E. in Accounting from Siena College. She was the recipient of the NSF CAREER award.

back to top