Summer Research at Stevens Institute of Technology
Kevin Lin
email: klin84@uclink.berkeley.edu
Advisor: Professor Yao/Jin
About Me
I am currently attending the University of California at Berkeley, having finished my first year. I am spending this summer doing research at Stevens Institute of Technology. I have lived in Parsippany, New Jersey, practically all my life. I enjoy tennis, skiing, and of course, programming.
Projects
Week 1:
Started working with java applets. Professor Yao wanted me to start with something I was familiar with, and then eventually work my way towards wireless-related research. So I made a TrigGraph and QuadGraph java applet, which, given parameters A, B, and C, would generate different graphs. I also worked with a Image Comparison applet, given quality and filesize, would select from a pre-processed image that would best correspond to that setting. Later, I worked on converting color to grayscale, and working with real-time processing of images. Went to the seminar Tuesday, and I listened to a professor talk about medical benefits with image processing.
Next Week's Plan: Continue working with the ImageComparison Applet, read up on edge detection, eventually working towards comparing filesize/compression ratio with edge detection deterioration.
Week 2:
Continued working on similar java applets. I looked into various edge detection techniques, and then made a java applet that would allow the user to compare and contrast various techniques. The applet allows the user to choose from a number of operators, including Simple (standard first order edge detection algorithm), Roberts Cross Operator, Prewitt Operator, and Sobel (3x3) Operator. Also, the user can select different levels of Gaussian Blur, Non-maximum suppression, and Thresholding levels. Also, I expanded the applet to incorporate handling several different images, as well as a path box for the user to input a custom image to be manipulated. I tried to display a file dialog box, but unfortunately, the code for that is incompatible with Microsoft Internet Explorer. There seems to be a bug in the Non-Maximum Suppression algorithm that I have yet to figure out.
Week 3:
I put the image processing algorithms aside, and Professor Yao and Jin instructed me to start doing some reading on wireless-related material. I read about smart antennas, and fading in a wireless environment. On Friday, I discussed with Professor Yao and Jin, and they suggested that I use the same graph applet to model wireless equations. I tweaked the graph applet, so that it could model several equations at once, display different colors, and have both an "overlay" and "side-by-side" mode. It user interface is still somewhat confusing, so I will continue to work on that. The website is now operational on the Steven's attila server.
Week 4:
After meeting with Professor Yao, we agreed that EnhancedGraph was probably too complicated for our needs. I drew up 5 different designs for a simplified version of EnhancedGraph (see SimpleGraph, above). On our Friday meeting, we agreed on SimpleGraph5, with a couple changes, which include a refresh button and pre-selected colors of red, green, and black. Lastly, I started implementing wireless related equations using the graph template. Given a list of equations and concepts from Professor Yao, I modelled the Shannon Capacity Equation using the same GUI was SimpleGraph5. The user selects a certain frequency (B), and the applet generates a corresponding equation using the formula C = Blog2(1 + gamma), where gamma is the Signal-to-Noise-Ratio (See Shannon Capacity Above). I also started working on the Galois Feedback Generator, which I will discuss with Jin today.
Week 5:
I worked on two applets this week: the Galois Feedback Generator and the Rayleigh, Rician, and Nakagami Fading PDFs. In order to model the PDFs, I had to implement the Zero-Order Bessel Function as well as the Gamma Function. I also implemented the erfc and erf functions for future use. It was hard to use a series approximation to find accurate results for the Bessel Function, and I think I will implement a real-number class that will handle larger numbers. The Gamma Function and the erfc function reasonably accurate estimates. I will also work on the Rayleigh Fading Simulation, converting the Matlab code Jin gave me into code that can implemented using the same "SimpleGraph" template.
Week 6:
I worked on converting a MatLab program that Jin had written into a java applet. This code models the Suzuki Fading Model. I ran into some difficulties with converting the MatLab code into Java because MatLab supports complex numbers, whereas Java only supports real numbers. Using Jin's suggestion, I modeled a real number with one set of vectors, and complex numbers with another set of vectors, and did the appropriate operations. Also, the erf function proved too slow for purposes of constructing an applet, so since the numbers that required the erf function were constant, I stored the output of the function as another vector to speed up the operation. I also worked on modeling the Fibonacci Feedback Generator using the same technique as the Galois Feedback Generator. However, I still need to confirm the results with Jin and Professor Yao.
Week 7:
After talking with Jin, I figured out my error with the Fading PDFs. However, the sum of the Rician Fading Model still does not equal 1, so there is still an error. I worked on producing graphs describing the BER of different modulation schemes. The results from the AWGN fading were similar to the results in the book; however, the results of the Rayleigh Fading are odd. I will discuss with Jin and Professor Yao today and iron out the bugs.
Week 8:
This will probably be my last week at Stevens Institute of Technology. After talking to Jin, I have figured out the source of all the major bugs in my applets, and I have fixed them. I also combined the Fibonacci Sequence Generator and the Galois Sequence Generator. Lastly, I have written a final report on my work during the summer. I have also included several word documents detailing each class and how they work in the applet.