randomarray.m download randomarray.fig download
A MATLAB GUI platform for displaying the radiation pattern of a narrowband beamformer with user-defined array geometry. User can specify the array geometry, directions of incoming signals, noise level, and the types of beamformer by mouse clicks. Useful for gaining insight about collaborative beamforming in sensor networks and random arrays. You need both files to work.
makearray.m download makearray.fig download
A MATLAB GUI platform that allows user to deploy sensors in a 2-D field with mouse clicks. You need both files to work.
ECE345, Visual-to-Audio Electronic Travel Aid
Code for TM320C54x (v2a.asm) download
This project involves the design and implementation of a audio synthesis device that converts moving images into audio signals. The system is built on a TM320C54x DSP with interface to an IMAQ camera module via the serial port on a PC. Brief description: A LabVIEW VI acquires an image from the IMAQ camera module. It quantizes the image into a 5x5, 3-bit image, and sends the data to the TM320C54x DSP via a serial port. The TM320C54x DSP constructs a 64-tap FIR by combining a series of 64-tap head related transfer functions (HRTF) according to the incoming data, and then filters an input audio signal with this FIR filter, in effect creating a correspondence between the filtered signal and the original image.
slide.tex download
This is a template for making slides in LaTex. Tips: In VI, you can assign F12 to texify your file by typing
:map #12 :! latex slide.tex & dvips slide
at the beginning of each VI session. I use it all the time and I almost do not use the mouse at all. You can save 1/3 of your time if you avoid mouse and learn hot keys.
uiucthesis_2e.sty download
(NOTE: Outdated since 2008) The style file for UIUC thesis. Original author unknown. I modified it a little bit so that it solves the little indentation problem on the table of contents page.
ECE580, Spring 2005, Distributed Maximum Likelihood Localization for Wireless Sensor Network
In this project, we derive a distributed algorithm in a wireless sensor netowrk for computing the Maximum Likelihood estimate of the node locations based on the relative location measurements and the nominal location measurements. We show that the solution to the ML problem can be obtained by a simple iterative algorithm derived from the Contractive Mapping Theorem. We simulate the algorithm and show that the algorithm successfully localizes the network. We conclude that the calibration process using high quality relative location measurements can only reduce the location error up to a certain amount.
ECE497RB, Fall 2004, Blind Deconvolution using Feature Based Parameterization
A blind deconvolution method using feature-based parameterization on the signals is proposed. The signals are parameterized by a finite set of coefficients according to basis functions that exhibit certain features such as piecewise constant, bandlimited, etc. The set of basis functions can be chosen based on the prior knowledge about the physical source that generates the signal. After parameterization, blind decoloration is performed on the coefficients, and the problem becomes a simple bivariate quadratic optimization problem that is solvable and inexpensive to implement. With slight modification, the proposed method can be applied to continuous or high dimensional signal models. Applications such as blind equalization for BPSK signals and blind image restoration are simulated.
ECE318, Spring 2004, Video Denoising and Compression via Temperol Wavelet Shrinkage
Sample video clips
In this project, we investigate a video denoising algorithm that performs wavelet shrinkage in the temporal domain. The algorithm uses a modified 3-D raster scan method to reduce discontinuities when scanning a time-space volume of video data. The proposed method introduces phantom artifacts on moving objects, but achieves substantial noise removal on still objects.
ECE390, Spring 2003, Maximum Likelihood DOA estimation using the Conjugate Gradient Method
In this project, we explore the use of an unconstrained nonlinear optimization algorithm on a Maximum Likelihood (ML) parameter estimation problem. The parameter to be estimated is the direction-of-arrival (DOA) of a source signal impinging on an array of antennas. The cost function is the data-driven likelihood function. The objective of the algorithm is to search for the optimal DOA that maximizes the cost function.
ECE497MD, Spring 2002, A Bayesian Approach to Spatially Adaptive Wavlet Model Image Denoising
Spatially adaptive statistical modeling has been used frequently to represent wavelet coefficients for random signals or images. This model consists of an underlying variance field $\sigma_s^2$ that is unknown and needs to be estimated for effective denoising. In this paper we introduce a Bayesian approach by assuming the model parameter to be a discrete or continuous random variable with a prior distribution. The resulting denoising algorithm produces a Bayesian mixture of conditional MMSE estimates of the wavelet coefficients over all $\sigma_s^2$ values according to the data-driven posterior distribution. The proposed algorithm for both discrete and continuous cases are implemented and the result is compared to several existing denoising algorithms.
ECE312, Spring 2000, 5-stage IMIPS microprocessor LVL521
We designed a microprocessor that features 21 IMIPS instructions, 5-stage pipeline, 2nd level caches and branch prediction table. The processor resolves data dependancy and control hazard. Cycles per instruction for this processor is about 1.02. It outperformed all other groups in the class and won the class competition.
ECE291, Spring 1999, Picture Twister
We developed an image morphing algorithm using x86 assembly under the NASM environment. The program allows users to distort the image by rearranging the location of the pixels. More details can be found at http://www.ifp.uiuc.edu/~clam/pictt/
ECE320, Fall 1999, Implementation of Discrete Multitone Modulation Power Allocation Algorithm.
I implemented a power allocation optimization algorithm for DMT on a TI54x DSP. It assigns bitrates to subchannels of DMT such that the overall power consumption is minimized. Optimization is performed via a fast gradient search algorithm.