Radio Frequency Identification (RFID) technology is currently employed for a variety of applications such as RFID-based wireless payment, healthcare, homeland security, asset management, etc. Due to newer privacy requirements and increasingly secure applications, typical RFID tags are required to expand security features such as data encryption and safe transactions. However, RFID tags have extremely strict low-power consumption requirements. Thus, reduced power consumption and secure data transactions are two main problems for the next generation RFID tags. This dissertation presents an architectural approach to address these two main problems. This dissertation provides a multi-domain solution to improve the power consumption and security, while also reducing design time and verification time of the system. In particular, I describe (1)a smart buffering technique to allow a tag to remain in a standby mode until addressed, (2)a multi-layer, low-power technique that transcends the passive-transaction, physical, and data layers to provide secure transactions, (3) an FPGA-based traffic profiler system to generate traces of RFID communications for both tag verification and power analysis without the need of actual hardware, and (4) a design automation technique to create physical layer encoding and decoding blocks in hardware suitable for RFID tags. This dissertation presents four contributions: (1) As a result, based on a Markov Process energy model, the smart buffering technique is shown to reduce power consumption by 85% over a traditional active tag; (2) The multi-layer, low-power security technique provides protection against malicious reader attacks to disable the tag, to steal the information stored in or communicated to the device. The power consumption overhead for implementing these layers of security is increased approximately 13% over the basic tag controller; (3) In addition, the FPGA-based traffic profiler system has been able to generate traces for ISO 18000 part 6C (EPC Gen2) protocol; and (4) The designs of endocing/decoding blocks are generated automatically by the Physical Layer Synthesis tool for five protocols used in or related to RFID. Consequently, any power consumption of five designs is less than 5 £gW. Furthermore, compared with five designs implemented by hand, the difference of the power consumption between two of them is less than 7% at most.

There exists a class of systems with requirements for real-time data delivery, limits on end-to-end delay, and limits on jitter. These systems can have components distributed across a wide area. In addition, the components distributed across a wide area require that the arrival and departure of data occur synchronously. To support these classes of systems, the communication systems must be able to transmit the required information within a pre-determined window of time. Due to the synchronous nature and requirements of these classes of systems, they are referred to as being Synchronous Dependent (SD). This research models and characterizes a serial communication link for application in a strict time constraint environment. These applications will also have limitations on jitter and delay, relative to the need to synchronize with other components of the system. Additionally, the research provides the modeling of users that utilize applications with relaxed constraints. The communication link will be able to support multiple users with varying requirements, from highly periodic control data to aperiodic general data. The network link is a modified T-channel, with resource reservation applying to both bandwidth and size allocation of a data frame. In contrast to a standard T-channel, the link has adjustable channel sizes, as well as the capability to shift a transmission out of the assigned channel into a channel either earlier or later than the previously assigned channel. In addition, a user may use more than one consecutive channel for the transmission of a single instance of information, i.e., multiple channels can be viewed as concatenated for use by a single user. The purpose of the channel is to provide a dedicated time slot available to the user that needs to transmit at a specific time that is also periodic. Through the modeling of the user's communications across the link, it is possible to examine the potential effects of the various characteristics of the individual user on the other users requesting access to the link. Assuming no adverse affects and to insure that time sensitive data are delivered on time, a method to determine the acceptability of the admission of the given user has been designed to determine which users will have access to the link and those that will not.

The Arabic language has many different dialects, they must be identified before Automatic Speech Recognition can take place. This thesis examines the difficult task of properly identifying various Arabic dialects. We also present a novel design of an Arabic dialect identification system using Hidden Markov Models (HMM). Due to the similarities and the differences between Arabic dialects, we build a ergodic HMM that has two types of states; one of them represents the common sounds across Arabic dialects, while the other represents the unique sounds of the specific dialect. We tie the common states across all models since they share the same sounds. We focus only on two major dialects: Egyptian and the Gulf. An improved initialization process is used to achieve better Arabic dialect identification. Moreover, we utilize many different combinations of speech features related to MFCC such as time derivatives, energy, and the Shifted Delta Cepstra in training and testing the system. We present a detailed comparison of the performance of our Arabic dialect identification system using the different combinations. The best result of the Arabic dialect identification system is 96.67\% correct identification.

Palladium (Pd) single nanowires with less than 100 nm thickness have been successfully and repeatedly fabricated using electrophoresis deposition method. This work also demonstrates the use of a single Pd nanowire as a hydrogen sensor with extremely high sensitivity. Growth and sensing mechanisms are discussed in detail and an improved synthesis method is proposed and proved. Single Pd nanowires were electrodeposited within 100 nm wide Poly-Methyl-Methacrylate (PMMA) nanochannels by using a probe station. The PMMA channels were fabricated by Electron Beam Lithography (EBL) on a pre-patterned template fabricated via photolithography and lift-off processes. Through this method, nanowires were grown with widths ranging from 50 nm to 100 nm, and lengths from 3.5 ¦Ìm to 6.5 ¦Ìm. The nanowires, successfully integrated into hydrogen sensor devices, were able to sense hydrogen concentrations as low as 5 ppm (part per million) at room temperature. Three different nanowire structures were found, and the growth control of nanowire structure was conducted. Different sensing mechanisms were addressed in detail according to different nanowire structures. A newly developed gate assisted growth method were also proposed and approved in this work. Thinner and more uniform nanowires were synthesized with this method, and the growth time was greatly reduced. Experimental data were presented to approve the effects of gate voltage on nanowire diameter and growth time.

Tunable fiber Bragg gratings (FBGs) are key components for optical communications and sensing applications. Current tuning mechanisms include on-fiber electric heating, the piezo-electric effect, and mechanical stretching and bending. Unfortunately, all of these tuning mechanisms rely on external electrical power supplied by non-optical means. Additional electrical cabling increases manufacturing cost and the risks of failure associated with additional on-fiber electrical contacts and fragile packaging, which are susceptible to electromagnetic interference. These limitations make current fiber components no longer suitable for use in hostile environments, such as extreme temperature, corrosive, and humid environments. The research herein presents a tunable fiber Bragg grating device without sophisticated packaging and external electrical wiring. Shown for the first time, the resonance wavelength, spectrum width, and chirp can be directly controlled by in-fiber light as well as spectral responses of metal-coated fiber Bragg gratings. In-fiber diode laser light at 910-nm was leaked from the fiber and absorbed by the surrounding metallic coating to raise the gratings temperature and to change the background refractive index distribution of the gratings. Wide tunability of the resonance wavelength and spectral width was demonstrated in both uniform and linear chirped gratings. Applications of in-fiber light-powered active grating sensors are demonstrated for dual function temperature and flow sensors based on self-heated optical hot wire anemometry. A grating flow sensor has been experimentally evaluated for different grating lengths and input laser powers. The grating flow sensor demonstrated a minimum measurable flow velocity for nitrogen gas flow of 0.35-m/s at atmosphere pressure, which is comparable to or better than most MEMS-based flow sensors. Optical fiber is not used only for optical signal delivery, but also as a multi-function cable that can deliver optical power for on-fiber self-heating. This one-fiber solution provides a new dimension to designing multifunctional fiber sensors without compromising their intrinsic advantages, which include immunity to electromagnetic fields, low cost, long lifetime, and the capability to function in harsh environments.

In Radio Frequency Identification (RFID) applications, it is beneficial to know where in a three-dimensional space an RFID tag will operate with respect to the interrogating transmitter. It becomes a very complex problem containing numerous variables including transmitted power, antenna gains, orientation, etc. One well-known equation used to approximate the power that a tag can receive from an interrogating transmitter is the Friis Equation. However, the commonly used form of the Friis Equation contains assumptions that limit the validity to a single point, orientation, and polarization in space, which is usually the most favorable. These simplifications eliminate the reflection coefficients and polarization terms, and the gains lose their angular dependences. This dissertation will provide a mathematical model that describes the operation of a tag in the far-field from a more realistic perspective in a three-dimensional space. The complete form of the Friis equation will be used as the basic formulation to model the amount of power a tag can receive for any orientation at a given point in space. The dissertation will also include mathematical analyses of how the location of the data base station affects the performance of the system by applying the physics embodied in the complete Friis equation to the return transmission link from the tag to the data base station. The complete mathematical expression will be used to evaluate the performance of an RFID tag by depicting the three-dimensional area of operation. The functioning volume will be solved using the developed scaling factor method and will give an accurate portrayal of where a tag can be successfully read as a specified percentage of reads when all orientations and polarizations are examined.

ABSTRACT POINTWISE LINEAR QUADRATIC OPTIMAL CONTROL OF A TANDEM COLD ROLLING MILL John R. Pittner, PhD University of Pittsburgh, 2006 The tandem cold rolling of metal strip is a complex multivariable process whose control presents a significant engineering challenge. The present technology generally relies on a control structure wherein the interactive coupling between process variables is partially reduced by several single-input-single-output and single-input-multi-output control loops operating on certain variables to decompose the overall problem into several separate problems to attempt to allow independent adjustment of strip tension and thickness anywhere in the mill. However, while the existing systems generally have been successful in producing an acceptable product, their failure to completely counteract the effects of interactions between process variables has limited their capability for improvement in performance and in robustness to disturbances and uncertainties. Various techniques for improvement have been proposed and some have been implemented. Many of these techniques offer some improvements, but also have shortcomings. Therefore there is a need for a better approach. It is considered that the pointwise linear quadratic optimal control might fulfill this need. This dissertation investigates the theoretical and applied aspects of this technique for control of a tandem cold rolling mill using criteria based on practical applications. A mathematical model of the mill is developed from which nonlinear state space equations are derived. Using these equations, a pointwise linear quadratic optimal controller is developed, and its performance for variations in operating speed and product are confirmed by simulation. Robustness to disturbances and uncertainties is also confirmed by simulation. The results are compared with those using typical industry practice.

Monitoring head position in persons with a traumatic brain injury may provide a means for independent powered mobility. Given the often limited residual functions of attending, visual processing, and motor control, the operation of an electric-powered wheelchair must be constantly monitored to ensure the safety of these users. Human-directed support is not always available and does not encourage independent mobility. The solution proposed for this problem is placement of a magnet on the rear of the persons head. Strategically placed linear analog Hall effect sensors that are fixed in a stationary headrest can then track the magnet; thus, accurately determining head positioning. With this proposed head tracking, a specialized interface to the electric-powered wheelchair controller can be used to ensure the persons head is attending the direction of travel asserted by a conventional, direction-sensing joystick.

This work analyzes Continuous Automatic Speech Recognition (CSR) and in contrast to prior work, it shows that the CSR algorithms can be specified in a highly parallel form. Through use of the MATLAB software package, the parallelism is exploited to create a compact, vectorized algorithm that is able to execute the CSR task. After an in-depth analysis of the SPHINX 3 Large Vocabulary Continuous Speech Recognition (LVCSR) engine the major functional units were redesigned in the MATLAB environment, taking special effort to flatten the algorithms and restructure the data to allow for matrix-based computations. Performing this conversion resulted in reducing the original 14,000 lines of C++ code into less then 200 lines of highly-vectorized operations, substantially increasing the potential Instruction Line Parallelism of the system. Using this vector model as a baseline, a custom hardware system was then created that is capable of performing the speech recognition task in real-time on a Xilinx Virtex-4 FPGA device. Through the creation independent hardware engines for each stage of the speech recognition process, the throughput of each is maximized by customizing the logic to the specific task. Further, a unique architecture was designed that allows for the creation of a static data path throughout the hardware, effectively removing the need for complex bus arbitration in the system. By making using of shared memory resources and applying a token passing scheme to the system, both the data movement within the design as well as the amount of active data are continually minimized during run-time. These results provide a novel method for perform speech recognition in both hardware and software, helping to further the development of systems capable of recognizing human speech.

Position determination in RFID systems is a very useful aspect in tracking and managing items, to which a tag device is attached. Knowing the precise location of an item, relative to its environment, allows accurate mapping and control of the item with respect to those people, or things, that use the items services. Current methods employ zone based localization, triangulation, or tag response time calculation. These methods are not very accurate, and are all reader side solutions. A tag side solution has been developed at the University of Pittsburgh to precisely determine the position of a tag, or reader, using passive RFID technology. Using the electromagnetic field intensity created by a reader, a tag that passes through this field will experience more excitation when parallel to and at the center of the field. Translating this action into DC voltage, a bell shaped voltage curve is apparent. Identifying the peak of this curve, therefore, directly relates to identifying the moment when the tag and reader are parallel and centered. Transmission of an ID, or location, at that moment, then, precisely describes its position. With the described motion in a known, single dimension, the item can be controlled to an even greater degree. One application is that of unmanned transit control. Using multiple tags, a transit car can be slowed and precisely stopped at an approaching station platform. The research contained within proves the existence of the voltage peak, develops an algorithm to find the peak, and creates a tag that can transmit the position ID using harvested RF energy.