Connexions is a collaborative, community-driven approach to authoring, teaching, and learning that conveys the dynamic, continuum of knowledge. The project involves two basic, interrelated components: (1) architecture and software tools and (2) a Content Commons of collaboratively developed material. A global community of authors, using special authoring tools, continually converts "raw knowledge" into small, self-contained modules of information and places them in a repository, or Content Commons, to be used, re-used, updated, and adapted. Instructors can use a Course Composer software tool to weave modules into customized courses that can be placed on the Web, presented in class, or printed as paper text. Students and other learners access Web courses or individual modules directly using special visualization and navigational tools designed to highlight the non-linear "Connexions" among concepts both within the same course, but more importantly, across courses and disciplines. Unlike any other educational technology project of its kind, Connexions features a synergistic mix of both software and content, and our solutions involving community development, modularization, XML markup, editorial lenses, and intellectual property cater directly to the wants and needs of the academic community. The system currently hosts over 1,000 modules in the Content Commons and is used as the primary text for eight courses. Connexions is providing authors, instructors, and students with an environment for promoting collaboration and building communities to share knowledge.
ELEC 201 introduces undergraduates to the principles and practice of engineering design, and to engineering as a profession. Teams of three students design, construct, and program a small autonomous robot to engage in a competition at the end of the course. The course is completely self-contained, assumes no prerequisites, and is intended for all majors, all years. During the course the participants are exposed to issues that confront every practicing engineer, such as working within constraints, using available technology, design tradeoffs, iterative design, team dynamics, and meeting project specifications, milestones and time constraints.
In our theory of information processing, information is defined only with respect to the ultimate receiver. Consequently, no single objective measure can quantify the information a signal expresses. For example, this paragraph (presumably) means more to a signal processing researcher than it does to a Shakespearean scholar. To probe how well systems process information, we resort to calculating how well an informational change at the input is expressed in the output. Briefly, to quantify an informational change, we calculate a information-theoretic distance between the probability distributions characterizing the signals that encode two pieces of information. We assume the signals, but not the information, are stochastic. The Data Processing Theorem says that the distance between the outputs of any system responding to the two inputs must be less than or equal to the distance calculated at the input. In the research presented here, we use this framework to characterize how well likelihood ratio detectors and block decoders process the information encoded in their inputs.
In the quest for understanding the subsurface, geophysicists employ sound waves for imaging of layering and structures in the Earth. In order to obtain more insight in the reflection amplitude characteristics we have developed new techniques for visualizing and quantifying reflection patterns. The character of these reflection patterns is kept in the amplitude, radial frequency and geometry. As seismic processes involve data manipulation and storage of massive amounts of bits, there is a pressing need for efficient algorithms. Although multiresolution analysis with the discrete wavelet transform (DWT) is computationally efficient, it does not provide any geometry information. The novel complex wavelet transform that we have designed contains this type of information without loosing the beneficial properties of the DWT. The local signal parameters that we extract from the volume equivalent of the transform are the three-dimensional local scale, geometry and reflection strength. The local scale can be directly correlated with the stratification in the Earth, since it tells us about the local frequency of a reflection or series of reflections. The geometry is especially useful for the structural interpretation. It provides information on its lateral extension. Based on the geometric information that we extract from the data, we are able to construct a virtual topographic map of lateral data-slices. Combining this information with the local scale of the reflection patterns, creates the opportunity for bird's-eye view interpretation. Since scale and structure are strongly related, anomalies in these parameters are potentially an indication for hydrocarbon presence.
Selected applications of multicarrier modulation for communication systems.
The phase of a signal is typically defined as the phase of its Fourier components. Fourier phase has many drawbacks for representing edges in images. It is hard to model the changes in Fourier phase due to aliasing caused by downsampling or due to superposition of compact and isolated signals, such as multiple edges in images. We introduce here a new class of representations to deal with these issues. This new representation provides a localized description of phase that also reflects the smoothness of the underlying compact signal. We discuss how to deal with the problems of location uncertainty in edges by making use of this so called "envelope+phase" representation.
The ability to view video and images on cell phones has become attractive in recent times. However, most of the content is of large size and cannot be viewed on the small LCDs that cell phones carry. Two possibilities exist:Option 1 is simple to implement but is often not what the user wants. I will show how to abstract the basic features of an image through resampling and concentrate on two key implementation issues: algorithm complexity and memory efficiency. These issues gain importance because processors on cell phones typically run slow to save power and have very little memory (To cut manufacturing costs.) I will also show how Digital Zoom can be performed through the same algorithm.
- Show a part of the image that is of the LCD's size. (Keeps all the detail, even if it is a small block)
- Abstract the basic features of the image and display it. (Cuts out the detail, fits the whole image)
This work was done during an internship at Texas Instruments in connection with the Leadership University Program.
Edges are the dominant features in most images, with great importance both for perception and for compression. Most wavelet-based image coders, however, fail to accurately model the dependency among the many wavelet coefficients near edges. We introduce a compression scheme using wedgelets, which offer piecewise-linear approximations to edge contours. By extending the successful Space-Frequency Quantization (SFQ) algorithm for wavelet-based image coding, we provide a framework that finds the rate-distortion optimal use of wedgelets. The resulting method produces compressed images with improved visual quality and lower numerical distortion.
To bridge the mismatch between the sizes of images and display devices, we present an efficient and automatic algorithm to create an adaptive image representation called SmartNail. Given a digital image and rectangular display frame smaller than the image, we define the SmartNail as an appropriately cropped part of a suitably scaled-down image. We choose the SmartNail-defining parameters--down-scaling factor and cropping location--to maximize a bit-allocation-based cost function that quantifies the visual importance of the image content in the SmartNail. For JPEG 2000-encoded images, the SmartNail parameters can be determined using just the header information available in the encoded file. Hence only the wavelet coefficients required to reconstruct the SmartNail need to be decoded from the entire JPEG 2000 code stream. Consequently, the SmartNail construction requires minimal computations and memory requirements. Simulations demonstrate the effectiveness of SmartNail representations.
With traffic intense applications and large file transfers becoming more and more common on the Internet as well as on dedicated networks, inference of partial or global network states becomes crucially important. The hunger for bandwidth of streaming and high performance applications as well as the need for real time monitoring for network management, early detection of anomalies and verification of service level agreements have spurred the development of tools to measure and estimate unused networking capacity.In this poster, we present PATHCHIRP, our active probing algorithm which features innovative, exponentially spaced "chirp probing trains". The wide range of packet spacings used in chirps leverages the power of traditional back-to-back packet pairs while remaining overall light and non-intrusive. In addition, chirps are ideally suited for model-based inference which permits estimation of correlations, provides confidence intervals, prediction, and inference of a host of network relevant parameters.
PATHCHIRP is freeware, available in C+. In a DoE sponsored project we are collaborating with Stanford Linear Accelerator SLAC and Los Alamos National Lab to deploy this powerful tool on a large scale using the infrastructure of PingER.
Current wireless systems can support data rates in the order of 100 Mbps in indoor environments while 300 Kbps in outdoor environments. The goal of our research is to increase the data rates by an order of magnitude to enable wireless multimedia communications, and at significantly lower power levels to ensure longer battery life. In particular, we design transmission and reception schemes for practical systems which can support different levels of mobility. We achieve higher data rates through the design of channel independent signal constellations in the case of high mobility applications while our design for low mobility applications exploits the channel conditions via beamforming and power control techniques. The added advantage of these schemes is the lower complexity of receiver implementation in addition to the increase in data rates. Our solutions enable the integration of multiple antennas into future wireless systems in a very power-efficient and computation-efficient manner.
High flexibility, rapid evaluation and fast time-to-market are new challenges in emerging wireless communication designs. Traditional designs, focused on satisfying area-time-power constraints, had customized designs that were application-specific and had a long latency for evaluation and time-to-market. Current DSP designs, while providing flexibility, do not meet real-time constraints for these systems. We explore a programmable architecture design using a stream-based processor simulator that targets real-time performance for wireless systems. In addition, area and power efficiency are also obtained by targeting high functional unit efficiency and minimizing memory stalls. The robustness and scaling limits of the programmable architecture to rapid changes in the algorithm complexity and ever-increasing data rates are also investigated.
The low power communications group at Rice focuses on issues such as evaluating theoretical limits of multiple antenna communications in wideband regime and practical code design techniques for systems with non-linear amplifiers. We consider theoretical limits for dense multiple antenna systems where a large number of antennas are packed in a small area and thus independence of fading coefficients assumption is violated. Additionally, we look at how signal shaping and warping can be used in modulation and code design for next generation wireless systems. Other problems under consideration are the quantification of power savings through user cooperation and multi-hop communications. These techniques offer additional advantages in power savings and diversity using smart routing schemes.
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.
Past investigations into the formation of ad hoc networks and ad hoc routing have focused on computer simulations of mobile traffic. However, the proliferation of mobile electronic devices (mobile phones, PDAs, and laptop computers) combined with the commoditization of GPS chipsets has made realizable the construction of a large-scale testbed for both GPS and non-GPS aware ad hoc network applications. In this poster we present the Rice University Shuttle Bus Project (RUSH) as such a testbed in its first stages of development, in which we have outfitted the shuttle bus system at Rice University with custom wireless and GPS hardware.
In the fast paced world of wireless technology, the ability to rapidly design and prototype systems is not just valuable but essential to keep pace with innovation. We are attempting to harness the power of rapid prototyping by creating a wireless testbed system that is flexible and dynamic enough to adapt to future algorithms, such as space-time codes and multiuser schemes proposed for advanced communication receivers. Such a testbed would allow algorithm developers to evaluate their designs without the time and expense of conventional prototyping, and they would be able to significantly decrease the latency between ideas and implementation.In this poster we present a testbed containing an embedded architecture enabling cohesive multi-processor software-hardware co-design. The testbed consists of PCs, SignalMaster and Transceivers. SignalMaster is a heterogeneous platform from Lyr Signal Processing containing both a TI C67 floating point DSP, a Xilinx Field Programmable Gate Array (FPGA) and ADACMaster (fast D/A, A/D converter). It can be used as transmitter or receiver. We implemented the algorithms on general purpose and special purpose architectures and connected the system to RF hardware. Simulink Real Time Workshop is used to achieve high-level reconfigurability. It also allows different programs to be easily loaded and executed in real-time on the hardware platform. We are using baseband algorithms for channel estimation, detection, and decoding to study their performance.
The wireless testbed supports both WCDMA and IEEE 802.11b standards. Transmission over the wireless medium is being initially carried out using 900MHz radios from Linx Technologies and in the near future by using our custom 2.4 GHz radios. The performance statistics obtained from realistic channels can be compared with the results with simulated channels.
One of the predominant schools of thought in networking today is that monitoring and control of large scale networks is most practical at the edge of the network (host computers or edge routers). Identifying or estimating network topology and quality-of-service parameters is crucial to edge-based approaches. Our work focuses on Maximum Likelihood estimation methods for these important network identification problems. The methods are based on traffic measurements that can be easily made at the edge of the network without special-special purpose support from internal network elements (e.g., routers, switches).
The Dyadic Classification Trees and Dyadic Polynomial or Platelet Regression Trees presented in this poster illustrate both the similarities and differences between tree-based classification and regression (i.e., function estimation). Tree-based methods are widespread in a variety of applications such as medical imaging, pattern recognition, and data mining because they admit highly efficient statistical inference algorithms. Both the classification and regression approaches begin by recursively partitioning the input space in a tree-structured fashion, and selecting an appropriate subtree by pruning. Performance guarantees in the form of risk bounds are obtained for both methods. Differences arise in the exact forms of the objective functions and pruning algorithms. This contrasts with the traditional setup of CART (Classification and Regression Trees), where identical objective functions and pruning algorithms are used to grow both classification and regression trees.
In this poster we propose two distributed scheduling/MAC mechanisms for wireless Ad-Hoc networks. First we devise Distributed Priority Scheduling as a fundamental tool for QoS scheduling. In distributed Priority Scheduling the priority index of a head-of-line packet is piggy backed onto existing messages so that other nodes can better assess the relative priority of their head-of-line packet in relationship to others'. We devise a simple mechanism to incorporate this priority information into the IEEE 802.11 protocol and achieve most of the gains of an ideal schedule with only a moderate fraction of piggybacked messages. Second we introduce Opportunistic Auto Rate (OAR), an opportunistic media access protocol for multi-rate ad-hoc networks. With OAR, nodes with good channel conditions are granted access to the channel for a duration that allows multiple packet transmissions. Consequently by exploiting the inherent variations in channel conditions, OAR gains more throughput compared to state-of-the-are auto-rate protocols. Moreover OAR ensures that all nodes, regardless of their channel conditions, access the channel for a time-share equal to that achieved by single-rate IEEE 802.11 and hence maintaining the long-term temporal fairness properties of 802.11.
The Resilient Packet Ring (RPR) IEEE 802.17 standard is under development as a new high-speed backbone technology for metropolitan area networks. A key performance objective of RPR is to simultaneously achieve high utilization, spatial reuse, and fairness, an objective not achieved by current technologies such as SONET and Gigabit Ethernet or by legacy ring technologies such as FDDI. The core technical challenge for RPR is the design of a bandwidth allocation algorithm that dynamically achieves these properties. The difficulty is in the distributed nature of the problem, that upstream ring nodes must inject traffic at a rate according to congestion and fairness criteria downstream. Unfortunately, the proposed algorithms in the current draft standards have a number of critical limitations. For example, we show that in a two-flow two-link scenario with unbalanced and constant-rate traffic demand, a draft RPR algorithm will suffer from dramatic bandwidth oscillations within nearly the entire range of the link capacity. Moreover, such oscillations hinder spatial reuse and decrease throughput significantly. We are developing a new dynamic bandwidth allocation algorithm called Distributed Virtual-time Scheduling in Rings (DVSR). The key idea is for nodes to compute a simple lower bound of temporally and spatially aggregated virtual time using per-ingress counters of packet (byte) arrivals. We show that with this information propagated along the ring, each node can remotely approximate the ideal fair rate for its own traffic at each downstream link. Hence, DVSR flows rapidly converge to their ring-wide fair rates while maximizing spatial reuse. To evaluate DVSR, we are using a combination of theoretical analysis, simulation, and implementation, including a proof-of-concept prototype of DVSR on a 1 Gb/sec network processor testbed.
Networking server performance has improved substantially in recent years, due mostly to rapid developments in application and operating system level software and, to a lesser extent, improved hardware in the network interface. However, depending on the characteristics of the server workload, the CPU and/or the local interconnect can remain a performance bottleneck. We propose to use a flexible network interface card (NIC) to alleviate these bottlenecks. Using existing programmable NICs, we have implemented a proof-of-concept solution to show that we can dramatically reduce the amount of local interconnect traffic, consequently improving server performance. We also intend to show that programmable NICs can reduce the load on the CPU by offloading networking computations and providing a richer set of services to the operating system. However, the flexibility and memory capacity of existing NICs are limited, so we cannot test and evaluate these proposals simply by reprogramming the on-board processor. We are currently developing a flexible Gigabit Ethernet/PCI interface card to allow us to explore system designs that make extensive use of the network interface to improve networking server performance. Our network interface card includes a powerful FPGA and a DRAM .By implementing a programmable network processor in the FPGA, we will create a network interface that can be easily upgraded in software. This capability is important for us because it facilitates the deployment of new value-added services without hardware replacement. The Rice Gigabit Ethernet/PCI NIC will provide a full duplex Gigabit Ethernet interface and will be implemented for use in systems that support a PCI local bus. The on-board FPGA will be used to implement a programmable network processor, multiple DMA channels, the PCI interface, the DRAM interface, and the Gigabit Ethernet MAC.
Network interface data caching is a new technique to reduce local interconnect traffic on networking servers by caching frequently-requested content on a programmable network interface. The operating system on the host CPU determines which data to store in the cache and for which packets it should use data from the cache. To facilitate data reuse across multiple packets and connections, the cache only stores application-level response content (such as HTTP data), with application-level and networking headers generated by the host CPU. Network interface data caching can reduce PCI traffic by up to 57% on a prototype implementation of a uniprocessor web server. This traffic reduction results in up to 31% performance improvement, leading to a peak server throughput of 1571 Mb/s.
We describe a program that combines the power of DSP theory with the technology of high-speed, all-optical planar waveguide circuits to provide powerful, adaptable optical filtering and processing capability. The data path is all-optical without speed limitations.
A random laser consists of light scattering particles present in a gain medium. In such a system, lasing can occur due to multiple scattering without the presence of external feedback. We report our studies of random lasing in systems consisting of Rhodamine 640 perchlorate dye and dielectric nanoparticles of titania or silica. In addition, we examine the effect of adding resonant nanoshells to the mixture. The resonances of nanoshells can be tuned across the dye emission spectrum, and their effect on lasing threshold and emission spectrum are reported.
The study of the scattering of light and other types of waves has been a challenging research area for many years and extends over several different disciplines in science and engineering. Here, we describe the first measurements of the propagation of coherent, single-cycle pulses of terahertz radiation in a scattering medium. The propagating electric field consists of unscattered (ballistic) and multiply scattered (diffuse) waves, both of which can be detected. We measure the propagation constants for pulses in a dense collection of spherical scatterers, and compare to the predictions of the quasi-crystalline approximation. We demonstrate that the terahertz time-domain spectrometer is a new, useful tool to investigate scattering.
We present a new approach to laser induced photoacoustic spectroscopy in gases. A high-Q watch tuning fork was used as a sensitive element to detect and selectively amplify a resonant photoacoustic response. Feasibility experiments were conducted and a sensitivity of 1.2×10-7 cm-1W/√Hz was reached. Several configurations to enhance the photoacoustic signal will be described. Experimental results will be presented as well as details of further sensor development and applications.
The sheep often serves as an important animal model for the study of human cardiovascular system. Here, we develop a closed-loop mathematical model of its cardiovascular system. A distributed approach is taken in describing the systemic circulation, which is divided into cerebral, coronary, foreleg, thoracic, abdominal, and hind-limb circulations. Nonlinear aspects of the systemic venous system are described including nonlinear pressure-volume characteristics of small and large veins and pressure-operated valves in large veins. The complete integrated model mimics typical steady-state hemodynamic data in the supine position. It can also be used to predict the blood volume shifts and hemodynamic changes that accompany standing up. These include the short-term neurally mediated cardiovascular response to this orthostatic stress. Additional studies predict the circulatory response to an increased afterload (balloon inflation) presented to the right ventricle. This model is further used to predict the response of the ovine cardiovascular system to the implantation of the PAL (Para-corporeal Artificial Lung) device and to test the putative effectiveness of different PAL device designs.
We have developed a closed-loop model of human gas exchange which consists of gas transport in the lungs and in the tissue. The two gas exchangers are coupled via the circulatory loop in our previously developed human cardiopulmonary model. Thus, gas transport in the lungs and at the blood-tissue interface interact as part of a closed-loop system, where the output of one gas exchanger provides the input to the other. Also Included in the model are lumped descriptions of lymphatic flow, oxygen capacity of the myoglobin and peripheral chemoreceptor reflex control. The model predicts physiological variables that agree well with typically observed data in human under normal condition. Our simulations of asphyxia using this model suggest an important role for transport delay in maintaining the arterial O2 tension and resisting hypercapnia during asphyxia. The simulation results also indicate the diffusion limitation of O2 in tissue during asphyxia. With further refinement, it is likely that the model could describe various physiological states and help to better understand the biophysics of cardiopulmonary disease.
A model for the rat phrenic motor neuron (PMN) has been developed that shows many of the characteristics of PMNs. Using a single set of parameters, the model robustly replicates many experimentally observed behaviors of PMNs in response to pharmacological, ionic, and electrical perturbations.
This study is concerned with the development of a multiple compartment model of the isolated cerebral artery in rat. The smooth muscle/arterial wall complex is an important component of the circulatory model and serves as a "vasomotor organ", which provides the myogenic mechanism that underlies the phenomenon of the autoregulation of blood flow. We have focused on this myogenic mechanism and have developed a model fof the electrophysiological and contractile characteristics of the single smooth muscle cell of the posterior cerebral artery. This cell model is used to interrelate the topics of arterial wall stress, changes in transmembrane potential, intracellular CA2+ concentration and contraction. Moreover, the smooth muscle cell model is embedded in a larger arterial wall model which converts contractile activity into changes in lumen diameter. The complete model consisting of component models of cell, wall, vessel and testing apparatus is used to provide biophysically based explanations of the myogenic mechanisms underlying the autoregulation of cerebral blood flow.