We have been studying a number of newly-developed fullerene-based molecules in an attempt to further understand the mechanics of nanoscale motion and manipulation of molecular systems. With four C60 "wheels" connected to a central chassis structure, these systems are relatively stable but mobile on the surface. Our work has consisted of characterizing and manipulating these molecules and other similar derivatives in both UHV and ambient conditions at room temperature using the scanning tunneling microscope. Further derivatives of two and three freely rotating fullerene wheels are also under evaluation and show promise of interesting capabilities and characteristics under manipulation.
In this poster we present an analysis of complex nanostructures using the finite difference time domain (FDTD) method. The basic theory of the FDTD algorithm as well as specifics of our implementation is given. This method is then applied to metallic nanoshells subject to several different defects. These defects include bumps and pits in the shell surface as well as offset silica cores. The results show that, as long as the shell is continuous, the far field response nanoshells in the presence of bumps, pits and small core offsets is very robust. Lastly on this poster we demonstrate FDTD's ability to calculate the time evolution of both the electric fields and surface charge oscillations of complex nanostructures.
In the past few years, new methods for guiding electromagnetic (EM) energy in sub-wavelength structures have been of great interest to researchers in the Nanotechnology area. This has been driven by the strong motivation of miniaturizing optical devices to facilitate the fabrication of Nanophotonic devices and circuits, especially Nanoshells, which have been used in cancer prevention and treatment. The individual and collective behavior of these Nanoshells changes dramatically, depending on their shape, sizes, separating distance, and orientation in the complex environment of a given application. This forces changes in the scattering, absorption, resonance frequency and field strength values exhibited by the Nanoshells and it leads to structurally tunable Nanoshells. Our present work is focused on studying the behavior of individual Nanoshells. We are currently developing closed form expressions of extinction and resonance of electromagnetic fields in Nanoshells. These closed form expressions will consequently help drive new simplified models for Nanoshells. Some of the recent models developed by our group show a maximum error of 1 to 5 % for typical optical tuneability range of Nanoshells. Additionally these models provide simplified expressions for extinction, scattering and absorption cross-section and intuitively explain the optical properties of Nanoshells. Future work includes development of system model based on basic building block to the individual as well as the collective behavior of Nanoshells. This will help us in future development of efficient models of systems based on complex nanostructures.
T-ray reflection mode tomography is an imaging modality where a terahertz (THz) transceiver illuminates an object with T-rays at a set of different look angles and measures the back-reflected waves. Two new reflection mode imaging configurations for T-rays are presented that are analogous to the wide-beam reflection computed tomography used with ultrasound and the circular synthetic aperture radar used with microwaves. Using similar tomographic reconstruction algorithms to X-ray computed tomography, such as the filtered backprojection algorithm, images are formed of the object's edges. Reflection mode imaging has advantages in higher resolution and faster image acquisition time over transmission mode configurations. Due to the unique responses of materials to t-rays, these techniques could be applied to package inspection and quality control.
Sources and systems for far-infrared or terahertz (THz) radiation have received extensive attention in recent years, with applications in sensing, imaging and spectroscopy. THz radiation bridges the gap between the microwave and optical regimes and offers a great scientific and technological potential in many fields. However, wave guiding in this intermediate spectral region still remains a great challenge. Neither conventional metal waveguides for microwave radiation nor dielectric fibers for visible and near-infrared radiation can be used to guide THz waves over a long distance, due to the high loss from the finite conductivity of metals or the high absorption coefficient of dielectric materials in this spectral range. Furthermore, the extensive use of broadband pulses imposes an additional constraint of low dispersion, which is necessary for compatibility with spectroscopic applications. Here we show how a very simple waveguide, namely a bare metal wire, can be used to transport THz pulses with virtually no dispersion, extremely low attenuation, and with remarkable structural simplicity. As an example of this new waveguiding structure, we demonstrate the first endoscope for terahertz pulses.
The development of ultrafast laser technology has allowed easier access to terahertz radiation for spectroscopy, sensing, and imaging. This is most commonly accomplished by photo-exciting a micron length dipole antenna deposited onto a dielectric surface with ultrafast laser pulses. However, much difficulty has been encountered when attempting to couple this terahertz radiation into waveguides, especially in regard to dispersion effects and coupling efficiency. Effective waveguides would allow controlled propagation of terahertz radiation and be significant to many potential time-domain spectroscopy applications. Our present research effort simulates a terahertz antenna using finite element method (FEM) modeling. Initially, a linearly polarized terahertz emitter on Gallium Arsenide (GaAs) was modeled and compared to previous results. A novel radially polarized terahertz antenna was then designed and modeled on the same substrate. It is expected that the radially polarized terahertz beam from the final antenna design will, among other things, couple to a coaxial waveguide with significantly greater coupling efficiency than previous efforts. This will open terahertz spectroscopy to many new applications and also enhance its function in many previous applications.
We use computer simulations to investigate the feasibility of constructing a planar, integrated optical encoder/decoder system using complementary spectral codes. Such codes are particularly useful for optical code division multiple access (OCDMA) communication in a multi-user optical access network. We developed a technique to achieve true bipolar encoding and decoding for spectral OCDMA, and tested our system using a multi-user fiber optic test bed having encoders and decoders built using conventional optics, which are bulky, costly, and subject to misalignments. In order to determine the feasibility of constructing a planar, integrated optical encoder/decoder, we developed a computer model of the photonic chip, and used it to find the system performance as the parameters of the individual planar waveguide components are varied. The results show that bit error rates of less than 10E-9 are generally possible with current fabrication tolerances, but that code lengths longer than 24-bits require tighter tolerances for some device parameters.
Quantum corrections to the conductivity allow experimental assessment of electronic coherence in metals. We consider whether independent measurements of different corrections are quantitatively consistent, particularly in systems with spin-orbit or magnetic impurity scattering. We report weak localization and time dependent universal conductance fluctuation data in quasi-one- and two-dimensional AuPd wires between 2 and 20 K. The coherence lengths inferred from both methods are in excellent quantitative agreement, implying that precisely the same dephasing mechanisms are relevant to both corrections. However, preliminary data in quasi-1D Ag wires show quantitative disagreement below 10 K, a disagreement seen previously in quasi-2D Ag films. Possible explanations for this disagreement will be discussed.
Organic electronic devices have great potential as inexpensive and flexible components in low-cost displays, identification tags, logic for smart cards and etc. Progress towards this goal has intensified in the past few years due to dramatic improvements in device mobility, synthetic processes, film quality and better contacts to these materials. However, a thorough understanding of the nature of transport in these devices and the physics of contacts at work in these materials is crucial to further development of opto-electronic organic devices. Here, we present the progress we have made in understanding and characterizing the nature of transport and charge injection in solution-processed organic thin film transistors. The emphasis has been placed on nonlinear charge injection through extracting and modeling the contact current voltage characteristics over a broad range of temperatures and gate voltages.
It is now possible to fabricate transistors where the channel is a single small organic molecule less than 2 nanometers in length. Such devices are excellent tools for examining the physics of electronic conduction at the single-molecule scale. We have used single-molecule transistors to explore unusual collective electronic effects, in which a single electron on the molecule forms a quantum mechanical resonance with the electrons in the electrodes. The electronic properties also reflect the vibrational modes of the molecules in interesting ways. The result is a rich physics laboratory for exploring a number of fundamental issues that can have practical implications.
Rapid, in-situ measurements of trace-gases are of significance in a number of applications that include industrial process control such as monitoring of toxic pollutants in industrial exhaust gases (NO, NO2, CO and NH3 at parts per million (ppm) levels) or natural gas quality evaluation (COS, H2S at parts per billion (ppb) levels) as well as in environmental monitoring. One of the techniques, which can be used to perform this type of measurements, is mid-infrared laser absorption spectroscopy (MIRLAS). MIRLAS enables high sensitivity measurements of fundamental ro-vibrational absorption lines of molecular species in the spectral range between 3 and 20 μm. Application of a quantum-cascade (QC) laser as mid-IR sources allows the design of a sensitive, selective, compact, and liquid-nitrogen free MIRLAS trace-gas sensors. This work describes two sensor configurations employing a thermoelectrically cooled, distributed feedback (DFB) QC lasers, which meets the specific requirements defined by the mid-IR spectral characteristics of a particular gas species and the gas sampling method to be used. The first sensor architecture was developed for open path monitoring of NO at elevated temperatures in industrial exhaust gases. To detect NO, we selected an isolated and intense R(6.5) fundamental ro-vibrational transition at 1900.08 cm-1. This line exhibits the least interference from other species present in hot exhaust gases. However, the absorption by NH3, CO2, and H2 sub>O must be taken into account due to pressure line broadening. In order to capture and resolve the four overlapping lines it is necessary to use a sufficiently large QC-DFB laser wavelength tuning range and a narrow laser linewidth. This allows simultaneous concentration measurement for all four species. An applied fast wavelength scan consists of less than 10 sequential pulses with individually pre-set laser frequency positions, which allows the acquisition of a fast snapshot of potentially occurring turbulent fluctuations in industrial exhaust ducts. The second sensor architecture presented is dedicated to simultaneous COS and CO2 concentration measurements in the gas samples at reduced pressure using an astigmatic Herriott cell of 36 m optical path length. Such a sensor can be used to detect and quantify species relevant to natural gas quality monitoring. The QC-DFB laser used in this sensor configuration operates in a pulsed mode at 4.85 μm and can reach a number of strong absorption lines in the P branch of the COS fundamental ro-vibrational spectrum. A noise equivalent sensitivity (1σ) of 1.2 ppb was achieved by measuring a well isolated COS P(11) absorption line in ν3 band at 2057.6 cm -1 using laser pulse repetition of 125 kHz and 0.4 sec. acquisition time. To address the need for fast measurements high-speed data acquisition electronics and digital signal processing (DSP) technology capable of sampling and analyzing data at rates of 1 MHz have been developed for both systems.
In the current regime of subwavelength lithography, the reliable manufacturing of semiconductors is becoming more and more difficult. As demand for reduced feature sizes in Integrated Circuits continues to increase, lithographic technologies are increasingly unable to print circuits without distortion. Correcting for these distortions requires the intensive use of Resolution Enhancement Technologies (RET). These RETs are computationally intense and thus greatly increase the time required for chip fabrication, thereby lengthening the time to market of the devices. One such RET is known as Optical Proximity Correction (OPC) and adds small carefully designed sub-resolution assist features (SRAFs) as a final processing step before mask fabrication. In addition to computational increases, OPC has the effect of flattening out formerly hierarchical layouts, because all cells in the layout are corrected in a slightly different manner with SRAFs. The removal of layout hierarchy coupled with an increase in the number of vertices in the design causes a huge jump in data volume and difficulty in mask manufacturing. We investigate the reduction of fabrication time and data volume through intelligent methods of OPC. We present methods aimed at equalizing total OPC implementation time over the design phase and fabrication phase of the manufacturing cycle, rather then it being totally the burden of the fabrication stage. We intend to do this through a method that will restore some of the design hierarchy while reducing the computational time of OPC correction methodologies. Our methods for OPC will then be extended to other areas, such as design rule reduction, with the overall aim of reducing IC cost and time to market.
This research examines the hardware and software mechanisms necessary for an efficient programmable 10 Gigabit Ethernet network interface card. Network interface processing requires support for a large volume of frame data, low-latency access to frame metadata, and high computational requirements for frame processing. Our research proposes three mechanisms to improve programmable network interface efficiency. First, a partitioned memory organization enables low-latency access to control data and high-bandwidth access to frame contents from a high-capacity memory. Second, a novel distributed task-queue mechanism enables parallelization of frame processing across many low-frequency cores, while using software to maintain total frame ordering. Finally, the addition of two new atomic read-modify-write instructions reduces frame ordering overheads by 50%. Combining these hardware and software mechanisms enables a network interface card to saturate a full-duplex 10 Gbps Ethernet link by utilizing 6 processor cores and 4 banks of on-chip SRAM operating at 170 MHz, along with external 500 MHz GDDR DRAM.
This poster presents a mechanism that reduces computation and memory bandwidth requirements of TCP processing for server workloads. The operating system hands off established connections to network interfaces capable of handling TCP. While the interface takes over TCP processing, socket buffers remain in main memory, reducing the amount of memory required on the interface. The socket interface to the application remains unchanged. A prototype web server based on an existing programmable network interface card is used for performance analysis.
TCP Segmentation Offload (TSO) is a hardware modification that seeks to reduce the load on the host CPU by offloading the segmentation of TCP/IP packets. Typically, the operating system creates packets by dividing the data to be sent into small segments, appending headers to each of them, and then passing these packets out onto the network. However, TSO allows the operating system, with the help of the driver, to create a larger TSO frame of up to 64KB in size (instead of the usual 1.5K). This larger frame is passed to the card where it is broken into regular sized packets and delivered onto the network. TSO reduces the computational requirements of sending data on the network by both reducing computational tasks (such as segmentation and header creation) and improving the efficiency of the memory management in the operating system (such as memory/buffer allocation/de-allocation and page alignment). This modification improves CPU utilization for bulk TCP sends and in applications with a large computational and networking component (such as MPI, NFS, and web servers).
Developing high-quality passive components, especially spiral inductors, in both RF integrated circuit and SoC technologies is vital to realizing high-performance and low-noise integrated wireless systems. Spiral inductors continue to be a major roadblock to the automated design of analog systems. They suffer from complex loss mechanisms and consume large chip area, which lead to difficult characterization and expensive implementation. Designing spiral inductors remains a difficult and time-consuming task. To facilitate spiral inductor design automation, we have developed accurate and efficient closed-form analytical spiral inductor models that are orders of magnitude faster than conventional modeling approaches using traditional Partial Element Equivalent Circuit (PEEC)-based methods. Our models accurately consider the complex loss mechanisms that impact the inductor's overall quality factor, such as skin effect, proximity effect and substrate eddy currents. These models will facilitate the development of spiral inductor optimization and automated synthesis techniques that will improve overall reliability and time-to-market for integrated wireless systems.
When designing an embedded system, it is often necessary to write an emulation program to simulate the system. Conventionally, the emulator is written in a language such as C, which is ill-suited for describing digital hardware. The design of emulation software on such platforms is thus difficult, and cycle-accuracy is sacrificed for reduced implementation complexity. The present work describes a novel method for simulating embedded systems described in the Verilog Hardware Description Language. We present tools for simulating execution of a program on the ARM7 core, as well as an operating system framework written in Verilog.
Soft errors due to cosmic rays and alpha particles emitted from packaging materials will cause unexpected data upsets in nanoscale integrated circuits. The issue of soft errors and reliability in integrated circuits is a growing concern for mainstream applications. The unpredictable nature of soft errors makes the overheads of general fault detection and tolerance techniques unacceptably high for mainstream applications. We describe techniques for cost-effective reduction in the soft error failure rate in integrated circuits. Moderate costs with a with an order of magnitude reduction in soft error failure rates for sub-100nm process technologies are described. These techniques are easily incorporated into other design approaches that target power, area, and delay constraints.
Wireless IEEE 802.11 networks in residences, small businesses, and public "hot spots" typically encounter the wireline access link (DSL, cable modem, T1, etc.) as the slowest and most expensive part of the end-to-end path. Consequently, network architectures have been proposed that employ multiple wireless hops in route to and from the wired Internet. Unfortunately, use of current media access and transport protocols for such systems can result in severe unfairness and even starvation for flows that are an increasing number of hops away from a wired Internet entry point. Our objective is to study fairness and end-to-end performance in multihop wireless backhaul networks via the following methodology. First, we develop a formal reference model that characterizes objectives such as removing spatial bias (i.e., providing performance that is independent of the number of wireless hops to a wire) and maximizing spatial reuse. Second, we perform an extensive set of simulation experiments to quantify the impact of the key performance factors towards achieving these goals. For example, we study the roles of the MAC protocol, end-to-end congestion control, antenna technology, and traffic types. Next, we develop and study a distributed layer 2 fairness algorithm which targets to achieve the fairness of the reference model without modification to TCP. Finally, we study the critical relationship between fairness and aggregate throughput and in particular study the fairness-constrained system capacity of multihop wireless backhaul networks.
Recently, application developers are increasingly hosting their resource-intensive applications on a global-scale grid of data centers, connected using custom high-bandwidth links. Examples include e-commerce web-sites with significant amounts of dynamic-content, multi-player gaming portals, etc. This work concentrates on one such resource-intensive application, "dynamic-content web-sites" and designs a suite of three protocols for reducing the client delays in accessing the content. The protocols introduced minimize the client access-times despite the presence of widely-different traffic conditions like, "time-of-day" or diurnal variations in traffic, sudden traffic-bursts (flash-crowds) and Distributed Denial-of-Service (DDoS) attacks. We introduce two protocols: Wide-Area Redirection of Dynamic Content (WARD) and Server Migration to handle short-term bursts and long-term trends in traffic respectively. We further introduce a DoS resilient protocol which uses WARD and Server Migration to maintain the QoS requirements of non-malicious users even in the presence of malicious users. Thus, in this work, we propose a novel grid-architecture and validate our results through a combination of analytical models, simulation-based studies and testbed experiments.
We are building a multi-hop wireless network using current off-the-shelf IEEE 802.11 hardware as a proof-of-concept of the Transit Access Points (TAPs) Architecture. In addition, the network will provide a cost-effective broadband Internet service to the low-income communities of Southeast Houston through the non-profit, Technology for All (TFA), by eliminating expensive wire-laying costs. In the TFA Wireless Network, one wired connection would feed nine square miles of wireless coverage through the use of directional and omni-directional antennas, showing the value of the TAPs architecture. Fairness and throughput experiments will be ran over the network using layer 2 static rate limiters. Finally, real-world deployment experience will prove invaluable for the multi-hop wireless backhaul platform of custom hardware that is being developed by physical layer, media access layer and routing layer groups on the TAPs project using the Center for Multimedia Communication (CMC) Lab and extending throughout the Rice campus.
The design of low-power, small form-factor remote and mobile sensing systems has become a more feasible task in the past few years due to several continuing trends. The cost of solid-state sensors for a wide variety of applications keeps decreasing. Robust low-power and short-range radio hardware has emerged which can handle moderate to high data rates (approx. 1Mbit/s). Embedded microprocessors consume much less power than their predecessors while achieving much better levels of performance. All of these trends make feasible very dense networks of fixed and mobile wireless devices for use in many different sensing and decision-making systems. In this poster we present a low-cost hardware and software testbed, named GNOMES, that has been developed at Rice University to explore the properties of heterogeneous wireless sensor networks, In particular, we present various methods to extending the lifetime of individual nodes in the network, the design tradeoffs that this presents, and the impact that this will have on the performance of the sensor network.
Though several wavelet-based compression solutions for wireless sensor network measurements have been proposed, no such technique has yet taken into account the need to couple a wavelet transform tolerant of irregularly sampled data with the data transport protocol governing communications in the network. To this end, we present an irregular wavelet transform capable of adapting to an arbitrary, multiscale routing hierarchy. Inspired by the Haar wavelet in the regular setting, our wavelet basis forms a tight frame adapted to the structure of the network. We present results illustrating the approximation capabilities of such a transform and the clear reduction in communication cost when transmitting a snapshot of the network to an outside user.
Reconfigurable hardware is key for rapid prototyping and verification of wireless communication algorithms. Most wireless communication algorithms exhibit a high degree of complexity and parallelism. The programmability and inherent parallelism provided by Field Programmable Gate Arrays (FPGAs) make them well suited for prototyping this class of algorithms. In this project, we present work done on an FPGA-based IF transceiver architecture for rapid prototyping of wireless systems. The project is synergistic with the Rice Wireless Testbed initiative for providing the infrastructure required for implementing and testing cutting-edge algorithms in a controlled setting before their deployment. Low Density Parity Check (LDPC) decoding is presented as an application example.
Multiple antenna systems are an emerging strategy for designing wireless systems with high data rate and high spectral efficiency. The high degree of correlation between the multiple antennas translates into a more complex hardware design for the wireless receiver. Chip-level linear channel equalization based on LMMSE solution is proposed at the physical layer of the mobile handset. This solution is used to restore the users' orthogonality destroyed in high scattering Multiple Input Multiple Output (MIMO) wireless environments. In this project, we present a scalable, customized and flexible hardware implementation of channel equalization algorithms for WCDMA downlink transmission in 3G wireless systems. Optimized and power efficient Application Specific Instruction Set Processors (ASIPs) based on the Transport Triggered Architecture (TTA) are presented that can operate efficiently in broad range of channel environments (low/high scattering, and slow/fast fading channels).
Wirelessly linked networks of digital sensors, called a sensor network, present new opportunities to sense, compute, and discover at spatial resolutions that were previously infeasible. Here, we discuss the practical application of this technology to our project in monitoring high resolution climate data in the Peruvian Amazon rainforest. In order to maintain longevity with a severely limited power supply, we have designed the network to operate in a very low power state (averaging on the order of a milliwatt). We present the enclosure design, wireless capabilities, and results from an initial test. We also discuss future possibilities in RFID and acoustic tracking.
Simulation is an established design mechanism that allows the system designer to rapidly prototype the design space and evaluate overall system performance. Most simulation environments, however, do not properly handle asynchronous events, and have a notion of global machine state which not only limits system flexibility and configurability, but complicates timing accuracy and presents the user with significant software engineering overhead. In this research we present extensions to the Spinach simulation environment for network interface architectures and embedded systems in general. Specifically, we introduce simulation models for heterogeneous system-on-a-chip type devices that are based around the TMS320C6x series of digital signal processors from Texas Instruments, as well as MIPS R4000 based microcontrollers and hardware based coprocessors with customizable instruction sets. In addition to providing models for other system components common to many embedded systems (memory arbiters, routers, multiported memory states, cache and dram controllers), the simulation environment supports binaries compiled within the Texas Instruments Code Composer Studio in a bit true, cycle accurate manner. In doing so, we show the computational benefits of using runtime reconfigurable architectures on task based workloads in the embedded domain. Additionally, workload partitioning with FPGA based coprocessors and the inherent hardware-software cosimulation of these devices is discussed.
Evaluating ad~hoc network routing protocols is difficult due to the complexity of possible network topology changes and the resulting protocol interactions. The most common method of evaluation, network simulation, allows repeatable experiments but may fail to capture the precise behavior of the real system. On the other hand, testbed protocol implementation allows the real system itself to be measured but is much more time- and equipment-intensive and is generally much more difficult. To address this conflict between simulation and testbed implementation, in this work, we present the design and implementation of a new system that allows existing simulation models of ad~hoc network routing protocols to be used without modification, to create a testbed implementation of the same protocol. We have evaluated the simplicity and portability of our approach across multiple protocols and multiple operating systems through example implementations in our architecture of the DSR and AODV routing protocols in FreeBSD and Linux using the existing, unmodified ns-2 simulation models. We also illustrate the ability of the resulting protocol implementations to handle real, demanding applications by presenting a demonstration of this DSR implementation transmitting real-time video over a multihop mobile ad~hoc network including mobile robots being remotely operated based on the transmitted video stream.
Existing wireless networks are limited by the cost of the wired backbone. We propose that the conventional, wired infrastructure be replaced by wireless multi-hopping with sparse wired connectivity. However, achieving throughput comparable to wired networks over band-limited wireless channels requires an emphasis on spectral efficiency. We use multiple transmit and receive antennas to dramatically increase spectral efficiency by combining innovative feedback-based power control techniques with new MIMO constellation and code designs. Feedback allows us to achieve full spatial multiplexing with non-zero diversity, and our constellations are many dBs closer to the channel capacity than commonly used MIMO signal sets.
This poster describes the custom hardware design of a Transit Access Point (TAP). A TAP is a wireless network node equipped with multiple air interfaces, capable of providing high-speed data links to both mobile users and other TAPs. In particular, these TAP-to-TAP links allow an access point to be connected to a larger network without a wired link. The elimination of the need for a wired connection at every access point will significantly reduce the cost and ease the installation of additional access points. The multiple antenna wireless algorithms are very complex and resource intensive, necessitating this custom hardware platform. We discuss here the required hardware capabilities, high level design decisions and the resulting PCB designs.
Burstiness constitutes a complex dynamic character of network traffic which can aversely affect network performance. Being only little understood, modeling and analyzing traffic bursts form an important issue. We provide evidence that only few connections cause bursts, typically connections over short paths. We call them "alpha connections". The "alpha traffic" is the traffic which is generated by alpha connections and "beta traffic" is the residual traffic. We contrast the effects on a network queue of two models for the alpha traffic, a self-similar stable and high rate ON/OFF model. We conclude in strikingly different predictions for queuing behavior in settings corresponding to different what-if-scenarios.
Translation Invariant (TI) image denoising is a frame denoising method. It outperforms orthogonal wavelet thresholding by averaging a collection of image estimates in different orthogonal bases. In this poster, we propose local models for characterizing the statistics (bias and variance) of this collection of estimates. Using these models, TI's gain can be analyzed due to the convexity of the error metric (MSE). Motivated by the edge geometry and analysis of smooth regions, we design a special way to choose an estimate (accordingly the basis generating it) from the collection at each pixel. In this way, the visual quality improvement of TI can also be explained. Insights drawn from this perspective include: a) the mechanisms by which TI achieves gain are different in smooth and edge regions, and b) most gain comes from edge regions. We also point to an improved way of exploiting the statistics mentioned above, if the position information of edges is available.
The ear is an exquisite and complex instrument for analyzing sounds, one that provides a remarkable ability to hear in complex, noisy environments. Unfortunately, hearing impairment can dramatically reduce a listener's ability to understand speech, especially in adverse environments. Although traditional hearing aids can improve speech intelligibility in quiet, they do not compensate for listeners' loss of ability in noise. In an attempt to overcome this limitation, we approach designing hearing aid processing algorithms in a new way. Our strategy is to quantitatively measure and then computationally minimize the distortion present in model responses of the first neurons in the auditory pathway, the auditory nerve fibers. This computational approach allows us to objectively optimize hearing aid performance under a wide variety of realistic conditions inexpensively. We will discuss the implications of our findings on the next generations of hearing aid processing.