Sponsored by:
NSF logo
Contact Information:
Oakland University

2002 Projects

Advisor: Dr. Sethi
The overall goal of the project was to group images into semantic categories using shape information. The shape information was a line drawing obtained by using edge analysis techniques on the picture. From these line drawings, local descriptors would be used to categorize the picture. The emphasis of the project ended up focusing on developing a method for obtaining these local descriptors. Ideally, a vector quantization method would be used to group vertical lines together, horizontal lines together, and so forth. And these groupings would be invariant to translation. The hope was to use these local descriptors for content-based image retrieval. By representing the original images as matrices of indices to the generated codebook, images can be separated into groups using co-occurrence matrices or frequency histograms. Some initial trials using these approaches accurately separated six images into two groups.
Advisor: Dr. Mili
Most safety critical systems involve a human in the loop making critical decisions, often while being under stress. It is important to understand the impact of interface design on the users' decision-making processes and decision performance when the users are under realistic conditions of operation. The goal was to analyze the effect of stress on various characteristics of human decision-making, and then to use the results as a guide for interface design.
Advisor: Dr. Hoda
A four-to-one analog multiplexer was designed as part of an electronic nose. The electronic nose is a sensing device that can detect traces of predefined gases or levels of toxicity. The 4-to-1 multiplexer was needed to preprocess the signal coming from the sensors and guide them to the appropriate signal processing units. The multiplexer needed to minimize the complexity of the circuit and to have no signal distortion. A 4-to-1 analog multiplexer was developed that operates in the linear range of its transfer characteristics and gives a linear range that is 48-51% of the total input range. It can be implemented using 0.35 um technology, with a time delay of only 0.5-0.6 ns, and with maximum frequency limitations for the input signal of 4-6 GHz.
Advisor: Dr. Djamel
The long term goal of this project is to build a device that will aid disabled persons (those unable to use their arms, or unable to speak but can see!) to access a computer system on a daily basis. A digital camera will be used to track eye gaze direction, allowing the eyes to act as a mouse. This project focused on determining the location of a person's eyes in a digital image. With a simple skin detection algorithm and some image processing techniques the potential eye regions are minimized. The eye regions are determined based on the amount of overlap each region has with every other region. This approach worked well for straight on face views without a lot of background noise.