Poster Titles
in progress
ABSTRACT
Dengue causes an estimated 100 million cases annually, with 25,000 of these resulting in death. Discovery of an effective vaccine against all four dengue viruses has been hampered by the twin challenges of immunodominance and heterologous immunity. We develop a statistical mechanics model of T cell immunity. From the insights we gain by the simulation, we propose polytopic injecting as one promising technology. We also explore the possibility of creating a multi-epitope vaccine that recognizes each of four dengue viruses by selecting the appropriate epitope from each strain. We also utilize an immunological and epidemiological model to study the mitigation strategies for H5N1 bird flu, potential pandemic influenza virus.
Cochlear outer hair cells (OHCs) are cylindrical sensory cells that undergo axial deformations up to 5% in response to changes in transmembrane potential. This process, termed electromotility, serves as a feedback mechanism in the auditory system that enhances the sensitivity and frequency selectivity of mammalian hearing. The molecular mechanism of electromotility is not fully understood; however, conformational changes in prestin, a transmembrane protein crucial for electromotility, likely play an important role. Since membrane proteins are very sensitive to their environment, studies of the OHC plasma membrane are central to fully understanding prestin function. Here we utilize fluorescence polarization microscopy (FPM) to measure the ensemble orientation of di-8-ANEPPS in the OHC membrane. This requires extending current, analytical FPM theory by deriving equations appropriate for a cylindrical as opposed to spherical cell. The results indicate that di-8-ANEPPS orients symmetrically about the membrane normal at 22.9° with respect to the plane of the membrane and demonstrate the feasibility of future FPM studies in OHCs.
Ultrasound imaging is a key modality in medical diagnosis. Compared to other imaging techniques, ultrasound imaging owes its great popularity to the fact that it is a safe and non-invasive procedure for visualizing the interior of the body. A major limitation of ultrasound imaging is that the image quality is much worse than for in CT or MRI scanning. In this work, we propose a speckle reduction technique for visualization enhancement and auto-segmentation improvement of ultrasound images. This method is designed to utilize the favorable denoising properties of two frequently used techniques: the sparsity and multiresolution properties of the wavelet, and the iterative edge enhancement feature of nonlinear diffusion. With fully exploited knowledge of speckle image models, the edges of images are detected using normalized wavelet modulus. Relying on this feature, both the envelope-detected speckle image and the log-compressed ultrasonic image can be directly processed by the algorithm without need for additional pre-processing. Speckle is suppressed by employing the iterative multiscale diffusion on the wavelet coefficients. We validate our method using synthetic speckle images and real ultrasonic images. Performance improvement over other despeckling filters is quantified in terms of noise suppression and edge preservation indices.
The goal of the Rice Atmospheric Information Network (RAIN) project is to develop a network of rugged, low-cost, battery-powered weather sensing nodes. The nodes will be connected by low-power laser beams to measure path-averaged raindrop size distribution, rainfall rate, and wind velocity. The same optical beams will be used for communication between nodes to form a sensor network that delivers data, via the Internet, to a central computer. The path-weighting function for such a laser sensor, that is, how the rain rate at different points along the laser path are represented in the detected signal, is a critical characteristic of the sensor system. We have examined the weighting function using simulations of a simple model.
Exhaled breath analysis for diagnostic and therapeutic monitoring is a relatively nascent field. Exhaled nitric oxide (NO) levels are altered in several respiratory diseases. Single breath NO measurements at a constant flow rate are adequate for monitoring conducting airway inflammation in asthma but are equivocal for monitoring peripheral inflammation in patients with chronic obstructive pulmonary disease (COPD). Partitioning the NO into alveolar and conducting airway regions may yield a clinically useful and non-invasive assessment of peripheral inflammation in COPD patients. Exhaled nitric oxide measurements are presently made using the chemiluminescence technique. Chemiluminescence achieves the required sensitivity but has several drawbacks, including frequent calibration, the generation and destruction of ozone, and high voltage. For this project, an NO sensor was developed based on integrated cavity output spectroscopy using a quantum cascade laser operating at 5.47 μm (1828 cm-1). Advantages of the new technology include no calibration or ozone, smaller size and lower initial and operating cost. The nitric oxide sensor is being used to measure NO in a randomized, controlled, dose-escalating trial of 40 patients with severe, stable COPD (30 drug, 10 control) at the Veterans Affairs Hospital (Houston, TX) to test the hypothesis that curcumin (diferuloylmethane) will decrease exhaled NO and other indices of inflammation in COPD patients. A custom breath collector was built to collect exhaled breath from patients at various constant flow rates (0.5 to 15 L/min). An iterative algorithm is used to determine alveolar and conducting airway NO from the NO levels at each flow rate.
An ultra-compact trace gas sensor platform based on photoacoustic spectroscopy using a quartz tuning fork (QTF) and components-off-the-shelf has been developed for sensor networks. Quartz-enhanced photoacoustic spectroscopy (QEPAS) provides a sensor solution which is significantly less in cost and size than most sensors capable of real time optical trace gas detection while preserving the high detection sensitivity and selectivity. Reduction of size (and weight) of optical trace gas sensors is important for many real-world applications, such as in-situ medical breath analysis and in monitoring of manned spacecraft environments. Low-cost sensors with low power consumption are particularly needed in applications of trace gas sensor networks, such as wide-area environmental air quality monitoring, semiconductor fabrication, process gas contamination, and source localization for security applications. This poster will report a robust, handheld, battery (and/or solar) powered sensor using a fiber coupled near-infrared (NIR) telecom components-off-the-shelf capable of reaching part-per-million (ppm) detection limits in real-time with minimal parts cost. The first prototype trace gas sensor footprint is 7.62 cm x 12.7 cm x 3.81 cm (3"x5"x1.5") and has a maximum power draw of less than 4W including laser excitation and cooling. The complete sensor has a mass of less than 1 kg (consisting of 2 circuit boards 100 g each; copper heat sink 100 g; diode laser, fiber components , reference cell and tuning fork module 300 g) without batteries. Two embedded low power microcontrollers (TI MSP430F1611) perform all control and processing functions which include real-time direct digital synthesis and filtering to produce low distortion, tunable modulation and demodulation waveforms, ambient temperature and pressure measurement, and line locking via laser current control and laser temperature proportional-integral-derivative (PID) control. The sensor exhibits QTF thermal noise (Johnson-Nyquist) limited behavior and demonstrates simultaneous temperature control and current tuning of a telecommunications diode laser at 1.5 microns.
The modeling and design of subwavelength waveguides and optical components is crucial for the realization of on-chip optical communication in future nanoscale integrated circuits. In this poster, we present several recent advances in the modeling and design of nanophotonic structures. We have developed a new RLC modeling technique for metallic nanoparticles, which could be used as a basic building block to develop an equivalent circuit model for plasmonic waveguides. For planar plasmonic waveguides, we have created an efficient full-vector-finite difference field solver leveraging Implicitly Restarted Arnoldi method to determine the propagation properties of the dominant modes. The method has low computational complexity and can be applied to accurately model complex geometries and structures with fast varying field profiles. To facilitate integrated subwavelength optical communication, we have also developed a technique to achieve tunable omnidirectional resonance in planar metallic microcavity structures, which can be used to realize tunable filters in the optical frequency range.
Given the increasing demand for wireless systems in system-on-chip technology, the modeling and optimization of low noise amplifiers (LNA) is vital for meeting power and performance requirements for fully integrated RF receivers. In this poster, we present an accurate and efficient modeling and optimization solution for LNAs in system-on-chip technology. We develop an analytical circuit model that captures the impact of integrated spiral inductor parasitics and transistor short channel effects. We then utilize deterministic numerical nonlinear constrained optimization to concurrently optimize the input and output impedance matching networks and transistor parameters to ensure that the synthesized amplifier design meets performance requirements and has passive component values that are suitable for on-chip integration. When the optimized LNAs are simulated using Cadence SpectreRF, our methodology yields significant improvement in noise figure and gain over the values obtained using equation-based design techniques. Leveraging the Normal Boundary Intersection method, we are able to generate the Pareto surfaces between LNA performance metrics in seconds.
This paper proposes the use of a frequency domain estimation algorithm to realize the broadband feedforward active noise control (ANC). The proposed method simultaneously realizes faster convergence of broadband ANC and an efficient block by block frequency domain estimation process. To configure an M-tap adaptive filter, this technique uses M-channel maximally decimated alias-free DFT frequency sampling filter (DFT-FSF) banks, which are equipped with 1-tap adaptive filter in each subband. Computer simulation results using the measured impulse response of an actual experimental setup show the effectiveness of the proposed method. The use of the proposed frequency domain estimation algorithm enables greater than 10 times faster convergence than the normalized least mean square algorithm, and at the same time, the delayless structure enables broadband noise attenuation.
Digital micromirror devices (DMD) have proven to be a commercially viable MEMS technology for the video/projector display market. Researchers are coming up with innovative ideas to exploit the properties of these unique devices. DMDs have carved out a totally new field of research on their own in terms of cameras. We propose to combine a microcontrolled mirror with a single optical sensor so that it can additionally acquire images, rather than merely adapt current camera technology to serve as an optical sensor. In this project, we describe the designing of a prototype for a revolutionary digital camera wherein the TI DMD chip is used as the micromirror array. We have developed a practical image/video camera based on this concept and realized it through the use of compressed sensing. Our design has additional desirable properties including scalable output bit stream, variable image resolutions and video frame rates.The object of interest is focused upon the micromirror array of DMD which has a pseudorandom pattern mapped onto it. The reflected light from either of the "ON" or "OFF" mirrors is collected on an optical sensor which in our case is a photodiode. Every pseudorandom pattern gives one coefficient (photovoltage) and using these coefficients and the random seed one can reconstruct the image.
The essential feature of compressed sensing lies in the fact that we measure (sample) the image/video fewer times than the number of actual pixels. This can significantly reduce the computation required for image/video acquisition/encoding. The salient features of our camera based on compressive imaging can be listed as: single detector, universality, encryption, robustness and progressivity, scalability and computational asymmetry. The single sensor allows us to extend imaging beyond the optical spectrum and to the hyperspectral imaging regime very cheaply and efficiently.
This poster proposes a novel image inpainting algorithm to recover the missing regions in images and video using a redundant quaternion wavelet transform. The proposed algorithm estimates the magnitude and phase of the quaternion coefficients by modeling their geometry and linearity. The missing blocks of pixels are recovered by iterative reconstruction of the image with the estimated quaternion magnitude and/or phase. The new algorithm achieves significant savings in computation and redundancy over competitive methods.
Visualization of high-dimensional data has been a major research topic for decades because informative visualization can be a powerful tool for discovery of cluster structure. By producing a two-dimensional spatially ordered quantization of higher-dimensional data, Self-Organizing Maps (SOMs, an unsupervised Artificial Neural Network learning paradigm) provide a rich layer of knowledge about the manifold, which can be extracted with explanatory visualization. SOM learning and visualization have been successfully used in data mining and modeling in many areas (http://www.cis.hut.fi/research/som-bibl).Available visualization schemes for SOMs usually portray the similarities of the neighboring SOM quantization prototype vectors either by showing the distance between them or by showing their receptive field sizes. While these schemes work well for relatively simple (low-dimensional, low- volume) data, they are often unsuccessful in capturing cluster boundaries of intricate, high-dimensional manifolds. One reason for this is the insufficient representation of data distribution among prototypes. We propose an original visualization that significantly augments the currently available methods. We use a Connectivity Matrix (CONN) which is derived from Delaunay triangulation by assigning weights to the edges of the Delaunay graph. The weights indicate the data distribution within the adjacent Voronoi cells. Visualization of CONN over the SOM thus expresses data topology and hence guides a more sophisticated detection of cluster boundaries than is possible with current approaches. This new capability is especially helpful for high-dimensional and large data sets with many meaningful clusters of widely varying statistics, including rare clusters such as in hyperspectral images of remote sensing sites or biological tissues, or genetic microarray data.
In this poster, we will overview the COMPASS architecture (Collaborative Multiscale Processing and Architecture for Sensor Networks), detailing the application programming interfaces (API's) which enable COMPASS to implement a wide variety of distributed sensor processing protocols from the sensor networking literature. We will also overview one such protocol, developed under the COMPASS umbrella, for distributed wavelet processing in sensor networks, and we will illustrate the implementation of this algorithm using the COMPASS API's.
Feedback is known to enlarge the capacity region of a multiuser channel. However, none of the current information theoretic analysis account for resource usage of the feedback link. In this work, we adopt a novel two-way formulation for multiuser channel, which jointly designs the uplink and downlink communication. By assuming that nodes are half-duplex (in time) and feedback shares resources with data, all resource usage is accurately accounted in the system. Our achievable rate region shows that feedback is beneficial only if the the channel is two-way, i.e, there is data to be sent in both directions. For some special cases, our achievable region matches the outer bound giving the capacity.
We propose a coding scheme for half-duplex estimate-and-forward relaying. First, we present an achievable rate for the estimate-and-forward protocol. Guided by the information theoretic scheme for the aforementioned rate, we then propose a practical coding scheme. The proposed scheme incorporates several features to reduce receiver complexity without compromising performance. Binary LDPC codes are used in the broadcast phase. Estimation is performed by scalar quantization of the received signal at the relay. Finally, a procedure similar to maximal ratio combining is used to aggregate direct and relayed signals at the destination. An important advantage of our scheme is that it does not require symbol-level source-relay synchronization. The codes outperform direct and two-hop channel capacities, as well as decode-and-forward relaying when the relay-destination link is strong.
We will present a framework for Medium Access Control (MAC) development and performance evaluation. The framework, developed on the Rice University Wireless Open-Access Research Platform (WARP), allows us to interface any MAC protocol with any custom physical layer (PHY), thereby providing a powerful research tool. MAC protocols for our framework are largely written in C using Xilinx Platform Studio and targeted to embedded PowerPC cores within the Xilinx Virtex II-Pro class of FPGAs. A key innovation is a flexible interface between PHY and the MAC capable of passing any user-defined parameter, thus enabling novel cross-layer research.
We will present a framework that allows for the control and observation of a network of arbitrary size. This framework was developed for use on the Rice University Wireless Open-Access Research Platform (WARP). It allows for massive scalability unlike current schemes that are too unwieldy to extend beyond a handful of nodes. The framework takes a high-level Xilinx System Generator model, integrates it into Xilinx Platform Studio, and then transfers data from the project to Matlab via ethernet. To showcase our framework, we will provide a demonstration in the form of a custom-built spectrum analyzer.
Feedback in multiple antenna systems has potential for increasing throughput and reducing probability of outage. Though feedback in communication systems has a long history, only very simplified feedback mechanisms are implemented in multiple-input multiple-output (MIMO) antenna systems. This paper presents an architecture for a MIMO system in which limited feedback is used for beamforming. By exploiting the structure of the beamforming codebook, the proposed architecture reduces computational requirements significantly, making feasible the efficient implementation of a real-time transmit beamforming and receive combining MIMO system. Hardware test results and simulation results are used to validate the proposed architecture.
City-wide wireless mesh networks are rapidly being planned for municipalities everywhere. In our poster, we discuss critical issues involved with city-wide mesh from experiences in deploying and managing a Houston mesh access network. Namely, we perform a measurement-based performance study of topology factors in wireless mesh networks, quantifying the impact of topology on mesh performance. Findings include that random networks require double the amount of mesh nodes to achieve performance objectives in comparison with regular grid networks. We additionally explore network management issues unique to mesh including aspects of traffic management, capacity planning and opportunistic protocols to improve mesh performance.
Multiple-Input Multiple-Output (MIMO) is one of the revolutionary physical layer technologies witnessed recently. The degrees of freedom offered by MIMO physical layer could be used to increase transmission rates, improve the link reliability, or reduce transmission range and hence enhance the spatial reuse. The tradeoffs between these MIMO strategies have attracted numerous research efforts. More specifically, the maximization of individual MIMO link capacity has witnessed intense interest motivated by the adoption of MIMO at the physical layer of many high-speed wireless systems. For example, the upcoming IEEE 802.11n standard will utilize MIMO to achieve physical layer rates up to 600 Mbps. However, it is well known that the throughput achieved by the network is considerably less than the data rates offered by the physical layer. One reason behind this degradation is the medium access control (MAC) overhead due to the exchange of the control messages, medium contention, collision resolution, etc. In this work, we present a CSMA-based medium access protocol that exploits MIMO not only to increase data rates, but also to alleviate the unfairness of CSMA in multi-hop wireless networks. This objective is achieved by making each link to use only a subset of the available antennas. Therefore, other links in its vicinity will be able to transmit their data using some of their antennas. Nodes determine the number of antennas to be used based on the information exchanged in the control packets. We call our protocol Carrier Sense Multiple Access with Antenna Selection (CSMA/AS). What distinguishes CSMA/AS is that it does not require the synchronization of all nodes to a common clock, and hence, no space-time signal processing is required to decode data belonging to interfering links. The overhead of the antenna information is low compared to existing protocols. Our results show that CSMA/AS outperforms the widely used CSMA/CA with multiple antennas in terms of both throughput and fairness
We develop and analyze a low-complexity cooperative protocol that significantly increases the average throughput of multi-hop upstream transmissions for wireless tree networks. This protocol exploits the broadcast nature of wireless networks where the communication channel is shared between multiple adjacent nodes within the interference range. For any upstream end-to-end flow in the tree, each intermediate node receives information from both one-hop and two-hop neighbors and transmits only sufficient information such that the next upstream one-hop neighbor will be able to decode the packet. This approach can be viewed as the generalization of the classical three node relay channel for end-to-end flows where each intermediate node becomes successively source, relay and destination. We show that our protocol dramatically outperforms the conventional scheme where intermediate nodes simply forward the packets hop by hop. At high signal-to-noise ratio, it yields approximatively 100 per cent throughput gain.
FlexNIC is a reconfigurable and programmable Gigabit Ethernet network interface card (NIC). It is an open platform meant for research and education into network interface design. The NIC is implemented on a commercial FPGA prototyping board that includes two Xilinx FPGAs, a Gigabit Ethernet interface, a PCI interface, and both SRAM and DRAM memories. The Xilinx Virtex-II Pro FPGA on the board also includes two embedded PowerPC processors. FlexNIC provides significant computation and storage resources that are largely unutilized when performing the basic tasks of a network interface. The remaining processing and storage resources are available to customize the behavior of FlexNIC. This capability and flexibility makes FlexNIC a valuable platform for research and education into current and future network interface architectures.
Virtual machine monitors (VMMs) such as Xen allow multiple virtual machines running on the same physical machine to share hardware resources. To support networking, such VMMs must virtualize the machine's network interfaces by presenting each virtual machine with a software interface that is multiplexed onto the actual physical NIC. The overhead of this software-based network virtualization severely limits network performance. Concurrent direct network access (CDNA) was created to eliminate the performance limits of software multiplexing by providing each virtual machine safe direct access to the network interface and performing network traffic multiplexing directly on NIC. Eliminating these software bottlenecks provides substantial returns. With a single guest operating system, CDNA increases transmit throughput by 31% and receive throughput by 121% compared to software-virtualized networking within the Xen virtual machine monitor. Furthermore, concurrent direct network access also maintains higher bandwidth as the number of guest operating systems is increased. Compared to the performance of a single guest operating system, network throughput degrades by at most 15% with 24 guest operating systems using CDNA, compared to more than a 40% degradation with Xen.
Mobile systems, such as mobile phones, now enjoy tremendous processing power and more than half of the world population own a mobile phone. Nevertheless, mobile systems have quite limited interfacing capability between the device and the user and the device and the environment. Extending the interfacing capability of mobile systems can enable them for more and better applications. For example, a mobile system-based health monitoring application can continually record user health information from sensors worn on the body for a digital health diary. The Rice Efficient Computing Group seeks to extend the interfacing capability of mobile systems with a WBAN. We have developed the Rice Orbit platform, and started to address the energy efficiency challenge of wireless interfaces in mobile systems.
With ever increasing capacity and dropping cost, Internet-capable mobile phones are predicted to be an important means to bridge the digital divide. Digital divide, by definition of the International Communication Union, refers to ``unequal access to information and communication technologies and the services they provide.'' The goal of this study is multiple. We wish to first better understand the digital divide in teenagers from different socioeconomic backgrounds; we wish to study the potential for Internet-capable mobile phones to bridge the digital divide; and eventually, we also want to identify design challenges/opportunities for mobile phones to achieve this potential. To achieve this goal, we conduct surveys and interviews to collect information regarding the ownership and usage of mobile phones and services by teenagers from different socioeconomic backgrounds. Analysis of the collected information sheds insight into the affordability and usability of existing mobile phones/services, and the technological and socioeconomic issues for them to provide high-speed Internet access and healthcare services for low-income people. Our findings will also help design future low-cost, energy-efficient, and user-friendly mobile phones/services relevant to people from low-income communities and developing countries.
This paper presents a DSP/FPGA hardware/software partitioning methodology for signal processing workloads. The example workload is the channel equalization and user-detection in HSDPA wireless standard for 3.5G mobile handsets. Channel equalization and user-detection is a major component of receiver baseband processing and requires strict adherence to real time deadlines. By intelligently exploring the embedded design space, this paper presents a hardware/software system-on-chip partitionings that utilizes both DSP and FPGA based coprocessors to meet and exceed the real time data rates determined by the HSDPA standard. Hardware and software partitioning strategies are discussed with respect to real time processing deadlines, while an SOC simulation toolset is presented as vehicle for prototyping embedded architectures.
Low Density Parity Check (LDPC) codes are one of the best error correcting codes that enable the future generations of wireless devices to achieve higher data rates. This poster presents a novel flexible decoder architecture for irregular LDPC codes. The decoder supports twelve combinations of code lengths -648,1296,1944 bits- and code rates-1/2,2/3,3/4,5/6- based on the IEEE 802.11n standard. All the codes correspond to a block-structured parity check matrix, in which the sub-blocks are either a shifted identity matrix or a zero matrix. A prototype of the LDPC decoder has been implemented and tested on a Xilinx FPGA and has been synthesized for ASIC.
Our goal is to propose high performance detection/decoding schemes that can support high data rates while keeping the complexity as low as possible. Complexity of optimum detection/decoding schemes in wireless receivers generally shows exponential behavior with the number of antennas and modulation order. Sphere detection has been analyzed recently to reduce detection complexity. We propose a novel architecture to implement sphere detection. This modified sphere detection, Dynamic Threshold Sphere Detection (DTSD), uses different complexity reduction techniques to increase the data rate and reduce the bit error rate. Moreover, iterative detection/decoding schemes have been shown to be able to achieve near capacity. We propose an architecture that utilizes a soft sphere detector along with an LDPC decoder interacting in an iterative structure. Soft sphere detection uses a search method where the specific search levels are bounded independently. The bounds are based on the distribution of the candidates found in each search level. Our approach results in area reduction and achieves significantly better performance than the previously implemented schemes while supporting higher throughput.