Project Title:

Breakthroughs in Channel Coding (without Bandwidth Expansion) for Next Generation Wireless

Principal Investigator:

Carl Nassar, Assistant Professor
Electrical Engineering
Colorado State University

Collaborating Companies:

Idris Communications

Company Representative:

In this proposal, we introduce an innovative transmitter design which, much like Ungerboeck's channel coding method some 20 years ago, enables receivers to benefit from channel coding performance benefits, without any cost in bandwidth expansion. Moreover, our transmitter, designed specifically with wireless fading channels in mind, also enables receivers to exploit frequency diversity benefits, thereby ensuring this channel coding method is particularly well suited to wireless environments, where such technology is needed the most. We call our novel channel coding design carrier interferometry (CI) channel coding. Successful completion of this research will transform carrier interformetry (CI) channel coding from the idea stage into a real world technology.

Project Title:

Real-Time 3D Navigation Using Video

Principal Investigator:

Christian Debrunner, Assistant Professor
William Hoff, Assistant Professor
Division of Engineering
Colorado School of Mines

Collaborating Company:

Peak Performance Technologies

Company Representative:

Mike Leigh, Manager Software Engineering

The goal of this project is to develop, implement, and test an algorithm for tracking human body motion in time-synchronized multi-camera video without the use of specialized markers. Current commercial motion tracking systems, such as those sold by the collaborating company, measure human body motion by tracking retroreflective markers attached to the subject and illuminated such that the markers can be easily detected. Attaching the markers is a tedious process, and a markerless motion tracking approach would greatly improve system usability. Other applications of the proposed algorithm include body tracking for human computer interfaces, surveillance, smart rooms, and telepresence. The goal of this project is to develop, implement, and test an algorithm for tracking human body motion in time-synchronized multi-camera video without the use of specialized markers. Current commercial motion tracking systems, such as those sold by the collaborating company, measure human body motion by tracking retroreflective markers attached to the subject and illuminated such that the markers can be easily detected. Attaching the markers is a tedious process, and a markerless motion tracking approach would greatly improve system usability. Other applications of the proposed algorithm include body tracking for human computer interfaces, surveillance, smart rooms, and telepresence.

Project Title:

An Architectural Approach to Building Reusable Business and System Models

Principal Investigator:

Robert France, Associate Professor
Sudipto Ghosh, Assistant Professor
Department of Computer Science
Dan Turk, Assistant Professor
Computer Information Systems
Colorado State University

Collaborating Companies:

Qwest Communications

Company Representative:

Pete Reinig

The objective of the proposed CASI research is to develop and apply a UML-based framework for creating, certifying and deploying reusable artifacts. The reusable artifacts include system and software models, expressed in the UML, that are linked to components in a reuse repository. The manner in which components and models are organized in the reuse repository is determined by an Enterprise Architecture (EA). An EA provides an integrated view of an organization's information systems (ISs). The manner in which an EA evolves is determined by a strategic architecture plan. The EA and the corresponding strategic plan defines the context in which system development projects, aimed at adding new business capabilities, are carried out. An effective reuse infrastructure must provide mechanisms that support Domain Engineering (DE). DE is concerned with the development and management of reusable artifacts. DE includes (1) analyzing existing development artifacts to identify potentially reusable artifacts, (2) developing and packaging artifacts for reuse, (3) certifying the artifacts, and (4) making the artifacts accessible to potential users.

The proposed research focuses on developing industrial-strength support for DE based on an organization's EA. Techniques for creating reusable UML models using (1) UML extension mechanisms in the form of profiles, (2) metamodels, and (3) the notion of roles, will be developed as part of this research. The research will also investigate the role an organization's EA can play in developing a reuse infrastructure. A reuse repository in which artifacts are organized around an EA is currently under development, and will be used to deploy the reusable artifacts developed using the techniques developed in this research. The feedback gained through use of the artifacts will be used to evolve the DE techniques.

Project Title:

Muscle Simulation for Musculoskeletal Mechanics Analysis

Principal Investigator:

Doug Smith, Assistant Professor
Division of Engineering
Colorado Schooll of Mines

Collaborating Companies:

Rocky Mountain Musculoskeletal Research Laboratory

Company Representative:

Scott Walker, Senior Project Engineer

A muscle simulation method is proposed to enhance the usefulness of current musculoskeletal kinematic analyses through muscle and joint force prediction. The proposed method is based on a nonlinear finite element simulation that will derive its geometry from images obtained via MRI procedures. This computer intensive approach allows for the consideration of complex constitutive behavior, muscle interaction and large deformation which are required to properly model muscle. Successful completion of this project will provide information needed to increase implant life while improving our understanding of joint mechanics. A plan is provided for our continued research to develop a complete muscle simulation software system.

Project Title:

Collision and Contact Modeling between Deformable Tissues for Surgical Simulation in Virtual Environments

Principal Investigator:

Min-Hyung Choi, Assistant Professor
Department of Computer Science and Engineering
University of Colorado at Denver

Collaborating Companies:

Touch of Life Technology

Company Representative:

Karl Reinig

We will develop a computational scheme for modeling collision and contact between human-body soft tissue structures and simulate their displacements and deformations. Surgical simulation in virtual environments holds great promise for training future surgeons in common techniques, as well as all surgeons in new methodologies. Such technologies would allow surgeons to practice techniques numerous times without the cost and limitations of cadaver based models or ethical problems of using real patients. Variations in technique could be explored on identical virtual models so that they could be evaluated effectively. Efficient and accurate computational models for soft tissue would play a core role in visualization and surgical simulation in virtual environments. The key to the successful simulation is to model the realistic behavior of deformable soft tissues when they are influenced by contact reaction force. Our investigation will be focused on determining contact regions and calculating reaction forces at appropriate nodes and elements within the contact regionsWe will develop a computational scheme for modeling collision and contact between human-body soft tissue structures and simulate their displacements and deformations. Surgical simulation in virtual environments holds great promise for training future surgeons in common techniques, as well as all surgeons in new methodologies. Such technologies would allow surgeons to practice techniques numerous times without the cost and limitations of cadaver based models or ethical problems of using real patients. Variations in technique could be explored on identical virtual models so that they could be evaluated effectively. Efficient and accurate computational models for soft tissue would play a core role in visualization and surgical simulation in virtual environments. The key to the successful simulation is to model the realistic behavior of deformable soft tissues when they are influenced by contact reaction force. Our investigation will be focused on determining contact regions and calculating reaction forces at appropriate nodes and elements within the contact regions

Project Title:

Data Mining Support in Client Relationship Management

Principal Investigator:

Junping Wang, Associate Professor
Department of Mathematical and Computer Sciences
Colorado School of Mines

Krys Cios, Professor

Department of Computer Sciences and Engineering
University of Colorado - Denver

Collaborating Company:

Empact Solutions

Company Representative:

Jim Baker, Chief Technology Officer

Many large companies, through a variety of formal service contract relationships, increasingly provide and consume information technology services to one another. Because of the services' critical business value, consumers, providers, and independent monitoring companies have an increasing demand to operationally monitor these services. The providers and, in some cases, the third party auditors also need to monitor the supporting resources. This results in a vast quantity of operational data from many different points of measurement. The operational data can be transformed into contract compliance data, by determining when violations have occurred against either expectations or formal service level agreements. This creates both a need and an opportunity for these large companies to extract inherent and often hidden knowledge from both data sets to better manage their on-going service contracts. This project will start with a synthetic service contract database, and use data mining techniques to (1) classify contracts by their likelihood of future non-compliance; (2) analyze associations between contracts so that in-time warnings on potential non-compliance can be issued when certain operational events have occurred; and (3) when a hand-built decision tree is available to classify operational events from multiple points of measurement by the services they impact, refine and validate the classification process in the decision tree by data mining techniques. The essential research issue in this project is to integrate classification with association analysis so that the results from one technique can be used to improve the other technique's analysis. Before a certain number of operational events with a contract have occurred, there might not be enough information to provide a decisive classification. However, we can use association analysis to predict highly likely forthcoming events, and include these predicted events in the classification process. In the meanwhile, association analysis can use classification models to analyze correlations within each type of contracts.

Project Title:

Subspace Methods for the De-mixing of Hyper-spectral Images

Principal Investigator:

Louis Scharf , Professor
Electrical & Computer Engineering
Colorado State University

Collaborating Company:

Data Fusion Corporation

Company Representative:

John Thomas, VP of Research and Development

Multi-spectral imaging is used to produce detailed geographic maps, wherein each pixel in the map contains an intensity distribution over a band of wavelengths. Thus the map my be sliced at various wavelengths to monitor the environment or classify features of a scene. The fundamental inverse problem in multi-spectral imaging consists of de-mixing the image spectrum, pixel by pixel, into the underlying the constituent materials that could have produced it. The de-mixed pixels are then used to classify the scene that produced the original image. We propose to develop adaptive subspace detectors and estimators for classifying scenes from de-mixed multi-spectral images. We will adapt the theory of tolerance intervals to the very important problem of controlling false alarms in multi-spectral classifiers. In collaboration with our supporting company, we will apply our theory of subspace hyper-spectral imaging to recorded multi-spectral images and use the findings to refine the theory. The results of this program will apply to environmental monitoring and defense surveillance from hyper-spectral images. Moreover, the basic methodologies we develop will be applicable to any vector image processing problem wherein pixels are vectors rather than scalar intensities.