Bhaskar Rao - US grants

This work involves the analysis and application of constant step- size adaptive recursive algorithms. The PI is applying the weak convergence theory pioneered by Kushner and incorporating prescaling to study the tracking behavior of constant step-size algorithms. These algorithms will then be applied to the problem of spectral line enhancement. This in turn has practical application to the problem of detecting harmonic signals in noise, which occurs in communications, signal processing, geophysics, military applications, etc. In the second year, he plans to generalize the adaptive recursive method and extend it to adaptive lattice filters which have been found to be quite useful in many signal processing applications.

Rao This research is studying novel algorithms for computing solutions to least squares problems with constraints, especially those problems that are underdetermined. The primary application is expected to be in Magnetoencephalography (MEG), a potentially new modality for the imaging of the brain. Both algorithm development and analysis are being considered, with MEG providing the forum for the testing and evaluation of these algorithms. It is expected that this research will greatly enhance our understanding of nonlinear iterative techniques.

CCR-9902961
Rao

THEORY, ALGORITHMS, AND APPLICATIONS OF SIGNAL PROCESSING WITH THE SPARSENESS CONSTRAINT

This research project will examine the theoretical and computational issues that arise in signal processing problems with the sparseness constraint in several important application domains. The research plan includes using majorization theory to develop and identify suitable diversity measures whose minimization leads to sparse solutions. Then, to minimize these measures, a new class of optimization algorithms will be developed, analyzed, and applied. Algorithms based on a factored representation for the gradient along with Affine Scaling Transformation (AST) based methods of interior point optimization theory will be the starting point of this work. To facilitate a more comprehensive understanding of the methods, and to develop methods robust to noise, a Bayesian framework will be employed. The important extension to the multiple measurement vector problem will be studied greatly expanding the range of applications. Learning algorithms will be developed to tune the required overcomplete dictionaries for specific application environments, thereby increasing their overall effectiveness. Theoretical and algorithmic development will be guided by the requirements of the applications. Particular attention will be given to the applications of signal representation and neuromagnetic imaging using Magnetoencephalography (MEG) (a potentially important new modality for the imaging of the brain).

The need for more bandwidth and capacity in wireless systems currently is the main culprit for the
great interest in the development of wireless communications systems operating at millimeter wave
frequencies and higher. The future needs of broad-band interactive services (1Gb/s) demand the
application of optical fiber feed networks for distribution of the radio signals to and from the antennas at
the various base stations. Fiber-optic technologies have reached the stage where insertions into various
commercial RF systems are being considered. Today, there are three main steps in the evolution of
RF/Photonics systems for wireless communications. The first step has been in the direction of using
photonics to slowly replace conventional RF components, such as, the coax that is used to interconnect
the antenna to the electronics. Optical fibers, in contrast to coaxial cable, provide a more ideal medium for
broadband RF communication systems. The light weight property of fibers, and its immunity from other
signal interference make them very critical in the development of future RF distribution systems. The
second, and more challenging step, is in the seamless integration of photonics and RF wireless circuits.
The challenge in this step is to use photonics and RF circuits as complementary systems and blend them
together. Finally, the third step is towards the development of optically coupled antennas. In this step the
aim is to eliminate the need of local oscillators, mixers, amplifiers and a host of other parts by directly
feeding an antenna through a fiber at millimeter wave frequencies.
Here, it is proposed that an array of RF modulator/photodetectors be integrated directly to an array
of antennas. This new RF/photonic antenna array system, with the appropriate space-time processing and
coding, will form a iosmart antennaln that can enhance network coverage, capacity, and quality. It is
envisioned that a large number of such RF/Photonic antenna elements could be networked together into a
star configuration, feeding in and out of a radio hub.
As a transmitter, the proposed optoelectronic device operates as a photodiode, while as a receiver
the device operates as an optical modulator. It has already been demonstrated that this dual function of a
semiconductor electroabsorption modulator and photodiode in the same device for duplex operation, can
occur, using bias control as a transmit/receive mode control. For full duplex operation, two
modulator/photodiode devices need to be incorporated in the each transceiver element.
We propose to directly drive a coplanar waveguide (CPW)-fed slot antenna by converting optical
power into microwave power and vice versa using these RF modulator/photodetectors. As a transmitter,
the CPW line is connected to the active surface of the photodetector, from which the microwave power
propagates to feed the radiating slot. The photodetector is fed via an optical fiber from beneath. When the
device functions as an optical modulator, the receive function can also be achieved. Preliminary results
for a single antenna show that a very good bandwidth and radiation patterns can be achieved.
It should be noted that these elements can be interconnected via the fiber to achieve summation,
mixing and other signal processing functions, at the antenna site or at a remote site. Some preliminary
results have been achieved in the area of multiple functionality for the optoelectronic components, such as
modulation, photodetection, self-biasing and RF frequency mixing. They have shown properties, such as
high bandwidth and high power, that are desirable for the antenna applications. A main emphasis here is
to further investigate the material and device designs for the optoelectronic component that can
incorporate into the smart antenna architecture.
The proposed approach will have significant impacts on wireless communication systems by
providing higher system bandwidth capacity and enhancing their reliability. It may lead to a new type of
long distance, broadband network infrastructure that supports transparent transport of optical signals.
Our team is formed to provide the expertise in the four key elements for this proposed research.
Our project will provide a good opportunity to train graduate and undergraduate students in one of the
most exciting interdisciplinary areas in science (RF, photonics, signal processing and communications).
The interactions between the researchers at the different institutions will be aided by the close
collaboration that exists between the members of the group.

The long-term goals of this project are to radically transform the ability of organizations that respond to man-made and natural disasters to gather, process, manage, use and disseminate information both within the emergency response agencies and to the general public. The project explores a multidisciplinary approach consisting of two interrelated research thrusts:
Scalable and robust information technology solutions to facilitate access to the right information, by the right individuals and organizations, at the right time, and
Social science research that investigates the distinctive nature of dynamic virtual organizations, and the social and cultural aspects of information sharing across organizations and individuals.

Research challenges addressed include mechanisms to: enable crisis responders to become rich sources of vital situational information; seamlessly collect data from heterogeneous sources; translate low-level noisy data into meaningful information that can be effectively used for damage assessment and situation awareness; facilitate information sharing and collective decision-making across emergent virtual organizations; and rapidly disseminate information in the form most useful to recipients. Close collaborations with multiple government agencies have been developed to test and validate research in live environments.

The project is expected to result in robust information systems that enable first responders to make well-informed and better decisions, to prioritize their response, and to focus on activities that have the highest potential to save lives and property. The resulting timely and effective response can contain or prevent secondary disasters, and reduce the resulting economic losses and social disruption during disasters.

The project will create new shared data sets for text, video and data mining. This will allow a larger scientific community to test algorithmic innovations against these field gathered data sets. Our community outreach programs will help generate greater awareness of the role of IT, stimulating new innovations as first responders interact more closely with researchers. Our educational programs will generate a better trained crisis management work force.

he central active challenge we are constantly addressing in daily life is to correctly assess the intent of others ('What is she trying to do? ...') and the import of sensory events ('What - good or bad - may happen now? ...') based on active perception ('It looks to me like she is trying to ...') and retrieved associations (''And she was the one who ...'). The corresponding problem for cognitive neuroscience is to identify, ideally from non-invasive brain activity recordings, those patterns of distributed brain activity that accompany and support active human cognition and behavior. This problem has two parts: First, -What patterns of distributed brain dynamics follow from, accompany, and predict specific world events and subject behavior? -To fully understand the experience and behavior of subjects in performing a given task, we must take into account both the import of each task event to the subject and the intent of each of behavioral event. These factors cannot be known directly, but they may be accurately guessed or inferred, in many cases, from detailed recordings of subject behavior and from the specific historical context in which each recorded environmental or behavioral event occurs. In the case of electroencephalographic (EEG) and/or magnetoencephalographic (MEG) signals recorded non-invasively from the human scalp, a second part of the problem remains -Which brain areas generate the identified signal patterns?'

The usual approach to analyzing electromagnetic scalp data has been to separate recorded events and behavior into a few simple categories, to average the recorded brain dynamics time locked to each event category, and then to apply physical inverse source estimation methods to scalp maps of peaks in the resulting averages. This project will explore using new machine learning methods, including advanced independent component analysis (ICA) and sparse Bayesian learning (SBL) methods, to jointly model the recorded task event, subject behavior, and brain dynamic data recorded in a complex learning task. The project has two goals: First, to identify patterns in unaveraged EEG and/or MEG data that reliably accompany subject behavior in specific contexts, and second to determine the exact areas of the subject's cortical mantle that locally synchonize their electromagnetic activities to produce the identified scalp patterns. If successful, the project will enhance the value of noninvasive electromagnetic brain imaging for identifying and measuring, with high temporal and spatial resolution, complex, distributed patterns of locally synchronous cortical activity that support active human cognition.

Abstract
This research examines theoretical, algorithmic, and computational issues that arise in signal processing problems where there is a need to compute sparse solutions. There are numerous signal-processing applications where sparsity constraint on the solution vector naturally arises. Brain imaging techniques such as MEG and EEG, sparse communication channels with large delay spread, high-resolution spectral analysis, direction of arrival estimation and compressed sensing are a few examples. The generalization and extension of the sparse Bayesian learning (SBL) techniques considered in this research will broaden the application domain and provide a very powerful complement to the existing maximum a posteriori (MAP) methods commonly used and in some cases even surpass them.
The investigators study extensions and generalizations of the sparse source recovery problem to greatly broaden the application domain. A key consideration in the work is developing a rigorous framework to deal with dependency in the sparsity framework. Motivated by applications with sparse but local structure, the research considers intra-vector dependency in the single measurement case, as well as intra-vector dependency as required in the multiple measurement contexts, among others. The research also includes the development of connections between multi-user communication theory and the sparse signal recovery problem to shed light on the stability with which sparse signal recovery is possible and to develop an understanding of the limits of suboptimal source recovery methods. To deal with non-stationary environments, the research develops on-line adaptive algorithms that exploit the inherent sparse structure of the application. The research also includes evaluation of the resulting algorithms in several important application domains.

Level of Effort Statement
At the recommended level of support, the PI and co-PI will make every attempt to meet the original scope and level of effort of the project.

Wireless communication networks continue to get more complex. Communication over unused TV bands (cognitive radios), self-organizing ad-hoc networks for military and emergency applications, or the heterogeneous networks envisioned in the next generation cellular (LTE advanced) systems all are clear evidence of this trend. As consumer demands, dependence on wireless services, and the complexity of networks continue to grow, there is a need to manage the resources efficiently without overburdening the network owner and end user. To address these challenges, this research involves significant and novel enhancements to the building blocks of modern communication systems.

The modeling, feedback, and cognitive techniques being studied in this project are essential for the development of robust and efficient wireless systems. The research project involves the following three parts: 1) Development of novel channel modeling methods that decompose the channel into a specular component and a diffuse component. A key consideration in this work is developing a rigorous framework for channel prediction and utilizing the insight to develop robust feedback based Multiple Input Multiple Output (MIMO) systems that degrade gracefully. 2) The development and analysis of feedback based multi-user MIMO-OFDM systems. This includes novel channel estimation and representations schemes, novel schemes for encoding sparsity, and performance analysis of reduced feedback MIMO-OFDM systems. 3) The development of advanced cognition at the physical layer. This includes waveform design in cognitive radios based on timing consideration and more comprehensive models for channels to provide awareness based on location, learning and memory.

This research project examines the theoretical, algorithmic, and computational issues that arise in compressed sensing (CS) and signal processing problems where there is a need to compute solutions to problems in which the solution vector has many zeros. In addition to the exciting compressed sensing area, this research will benefit numerous signal processing applications where the sparsity constraint on the solution vector naturally arises. Brain imaging techniques such as Magnetoencephalography (MEG) and Electroencephalography (EEG) are currently important examples. Sparse communication channels with large delay spread, high resolution spectral analysis, and direction of arrival estimation, are other important examples. An effective solution to this problem will have significant impact, by providing new and valuable tools to the practicing signal processing engineer. In addition, the tools will be of interest to researchers in cognitive science, neuroscience, and machine learning where sparsity issues naturally arise, such as sparse coding of signals in the brain or learning from data which is often assumed to lie on a low dimensional manifold.

This project provides a comprehensive and tighter integration of the compressed sensing field and multi-user information theory. This makes it possible to utilize the rich results available in network information theory which have been successfully applied to the implementation of communication systems. The theoretical tools necessary to enable this integration are being developed by the investigators. This research enables significant advances in both theory and practice in the CS field. The information theoretic insights are leveraged to provide insights on performance limits and guidance on practical CS-based system design. The implementation experience gained from communication systems will be translated to practical algorithm development and efficient CS-based system design.

The demand for wireless services and higher wireless throughput continues to grow exponentially. To meet this growth, massive multiple-input multiple-output (MIMO) has been identified as an enabling technology in next generation wireless systems. A challenge in realizing the vision is the estimation of the wireless channel between the transmitter and receiver as the number of transmitting antennas becomes large. The channel modeling and estimation challenge is addressed in this research for a variety of deployment scenarios; frequency division duplex (FDD) systems, time division duplex (TDD) systems, and distributed massive MIMO systems. In addition to having a significant impact on the theoretical foundations and algorithms relevant to next generation wireless systems, this research will involve several graduate students who will be trained in the latest wireless technology and also result in novel tools that have fundamental and wider import.

The channel modeling and estimation research includes the development of line-of-sight channel estimation via advanced sparse signal recovery algorithms like sparse Bayesian learning with the goal of reducing training overhead. The non-line-of-sight environment is considered from a novel dictionary learning perspective to enable low dimensional representations of the channel. These representations along with compressive channel learning will lead to the development of techniques that significantly reduce the feedback overhead for FDD systems. For TDD systems, the research involves the development of data-aided channel estimation techniques to improve channel estimates well beyond what is possible with pilot-only training. In addition, the research includes an in-depth study of the tradeoffs of distributed massive MIMO array design to develop insights necessary for selecting the optimal array configuration.

The project addresses the challenges of next-generation wireless communication systems. A key enabling technology for reliable and high-data-rate communication is the deployment of Multiple Input Multiple Output (MIMO) systems, which consist of multiple antennas for transmission and reception. With the use of the millimeter-wave (mmWave) frequencies in next-generation systems, the shorter wavelength enables deployment of many antennas in a small physical area, leading to massive MIMO systems. Massive MIMO systems, however, tend to have high complexity, high power consumption and high cost. This project seeks to do more with less: “Less” refers to limited hardware (fewer radio-frequency chains, one-bit analog to digital converters, etc.) and “more” to being able to extract the benefits (with minimal degradation) of massive MIMO systems by working around these hardware limitations. To do more with less, the project adopts a synergistic approach where innovations in system architecture and algorithms (model-based and data-driven) complement each other via judicious exploitation of structure (antenna array geometry and modeling) aided by powerful inference frameworks (sparse Bayesian learning and machine-learning techniques). The project will lead to state-of-the-art wireless communication systems that should help with maintaining US leadership in this important technology as well to train the next generation of researchers in this area of strategic importance.

To develop low-complexity, low-cost, next-generation mmWave massive MIMO systems, this project has two major components. One is to harness antenna array geometry, both for one-dimensional and two-dimensional arrays, for rich channel sensing with fewer sensors complemented by robust inference. A key aspect of this work is embedding a nested array into a massive MIMO architecture employing fewer radio-frequency chains. The rich sensing capability of the nested array is being maximally exploited using the sparse Bayesian learning method. The channel sensing is also complemented with enhanced channel models incorporating variable and unknown angular spreads. A further component is the use of learning through deep neural networks to compensate for nonlinearities introduced to reduce power and cost. Models complemented by learning as well as fully data-driven techniques are being developed that address the specific and unique needs of wireless systems, such as variable numbers of users and channel coherence.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

The project addresses challenges of next generation 6G wireless communication systems. For these systems, millimeter-wave (mmWave) and terahertz (THz) frequency bands that support wide bandwidth transmissions will play an important role in providing the advanced services envisioned of next generation systems. Due to the small wavelength, a key enabling technology for reliable and high data rate communication is the deployment of massive Multiple Input Multiple Output (MIMO) systems which consist of a very large number of antennas for transmission and reception. This allows for dense spatial sampling and use of spatial degrees of freedom for effective communication system design. However, the small form factor makes traditional radio-frequency (RF) circuitry design impractical due to circuit complexity, increased cost, and power consumption. These constraints lead to nonlinearities that call for developing nontraditional processing algorithms for which recently developed machine learning networks are suitable. Another challenge is the wireless channel which at these higher frequencies has significant path loss and varies in nature across different frequencies in the bands. To deal with the higher path loss there is a need for finding ways to enhance the quality of the channel, to which this project applies advanced channel morphing methods. The theoretical ideas resulting from the work will be supported with appropriate experimental work to lead to practically viable systems. The project will lead to state-of-the-art wireless communication systems that should help with maintaining leadership in wireless technology as well to train the next generation of researchers in this area of strategic importance.<br/><br/>To develop next generation mmWave and THz based massive multiple input multiple-output (MIMO) wireless communication systems using machine learning (ML) algorithms, this project has four major components. One is ML-based sparse channel modeling in severely constrained environments, i.e., limited sensing, limited number of measurements, limited precision, and system imperfections. This work combines domain knowledge with data driven techniques to deal with the nonlinearities and imperfections in the system. A second component is novel channel modeling using block-sparse techniques and development of associated model-based and ML-based inference algorithms. Block channel structure is not analytically tractable in two dimensions and calls for ML techniques to learn from data. A third component is incorporation of reconfigurable intelligent surfaces (RISs) for channel morphing to improve channel quality. A final component of this project is experimental work, channel sounding and ray tracing, to support, validate, and refine the theoretical models.<br/><br/>This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.