Sponsored by:
NSF logo
Contact Information:
Oakland University

2005 Projects

Advisor: Dr. Zohdy
Extraordinary advances have been made in computational biology and bioinformatics due to discoveries from genome data sequencing and clustering of human and model organisms. New research hold promise to better characterize biological structure and function and to enhance the descriptive and predictive power of modern biology. As biological systems are better understood, analogies are used to create novel computer paradigms. These paradigms can in turn be used to study biological systems. Motivated by successes of a previous REU project that contributed to using novel neural networks for modelling metabolic activities of Ecoli bacteria, bio-inspired computer paradigms are again used to study biological systems. This research employs hybrid neural networks (feed-forward, Hopfield and Elman) in the area of immunoinformatics.
Advisor: Dr. Sethi
While multimedia databases have grown at an astronomical rate, intelligent organization of such databases has not kept pace. One method of organizing information is to cluster the information into groups of similar items. Many similarity measures exist, and one such measure is chi-square similarity. The chi-square similarity measure makes use of a confidence value and this specification may be used like a focus dial of a telescope. This project will explore the use of the chi-square similarity measure to organize a color image database at different focus levels of clusters.
Advisor: Dr. Elhajj
Nursing students would benefit from realistic training scenarios. Since they cannot practice on humans, it becomes necessary to offer a best approximation simulation. This project focuses on the simulation of intubation process use, in particular nasaltracheal suctioning. Nasaltracheal suctioning involves inserting a tube through the patient's nose and down the trachea to clear out the patient's lung. To simulate this process for nursing student training a haptic joystick will be used to present realistic force feedback through the insertion process.
Advisor: Dr. Mili
Sensor networks consist of large sets of sensor nodes distributed geographically to monitor events of interest (e.g. atmospheric changes, appearance of specific objects) or collect data measurements (e.g. temperature, light, pressure) at different locations. Each sensor node generates data records at predefined time intervals. The sum of all the records generated can be seen as a distributed database used to monitor conditions and make decisions. For example, a sensor network monitoring the water quality of the Detroit River would consist of thousands of sensor nodes positioned in and around the river sensing different chemical concentrations in the water. In traditional sensor networks, processing and analysis of the information collected by the nodes is done in a centralized location. This approach is highly inefficient because it uses scarce (battery) power to transmit data, most of which will never be used. This approach is ill adapted for most applications using sensor networks. This project focused on optimizing the execution of queries submitted to the network (e.g. Show me the highest concentration of toxic material and its location). Cost functions were developed to utilize different strategies for distributing processing amongst local nodes and to minimize the amount of data that needs to be transmitted in order to optimize query execution.
Advisor: Dr. Kim
Domain-specific design patterns provide a proven solution for recurring design problems in the development of software systems in a specific domain (the banking domain, the telecommunication domain, etc). One of the uses of domain-specific design patterns is to use them as design templates to generate similar software designs, which facilitates the development of high-quality systems in short time. The generated design is served as an initial design of the system being built and completed by the designer. In this project, students will design and implement a prototype tool that can systematically generate software designs from a domain-specific design pattern. Students will be exposed to the Unified Modeling Language (UML), a popular software modeling language, and C++ as an implementation language.