EEL6502: Projects

Project report due Monday April 8 in class at 3pm. You must email me a description of your proposed project by Monday, March 18.

Everyone must complete a project. The project will consist of 10% of your project, roughly the worth of two homeworks. Your project report should be written as if it were a paper to be submitted to a conference and therefore should contain the following components:

1. A short review about the topic, include any references that you use.
2. A concise description of the problem.
3. A detailed description of your solution to the problem.
4. Matlab simulation results.
5. A discussion of the significance of these results.
6. The appendix should contain complete MATLAB codes, messy derivations and any other information too detailed to keep in the main body.

You are strongly encouraged to come up with your own idea for a project. The best projects relate to something you have already worked on or familiar with. If you have the capability of measuring data of any kind yourself, then this can lead to an excellent project.

1. In the application chapters of Widrow's book, there are sections on dsp filter design, control, communications and array beam forming. Any applications in these areas can be used in projects.
2. Add varying amounts of noise to a signal and use a reference to recover the original signal the best you can. If you know the original signal, you can measure exactly how well your algorithm is working. The original signal can be speech, an eeg trace or any other real signal.
3. Adaptive noise cancellation without a reference signal.
4. Recover a periodic signal from broadband noise.
5. Time delay estimation.
6. Problems dealing with multiple time series data.
7. How many taps? Use the information theoretic criterion described in Haykin section 2.10 and develop an algorithm for choosing an appropriate number of taps. How well does it work?
8. So far we have assume that all signals are stationary. How quickly can the statistics of the signal change while still allowing LMS to keep up?
9. Linear prediction is a very popular area. There is a whole chapter of Haykin deals with linear prediction. You can predicting a whole host of standard time series data that are available on the internet, e.g. financial time series, chaotic time series.
10. Any theoretical analysis of convergence of LMS or other aspect of adaptive filters makes a good project. Make sure that you are the type of person cut out for this.
11. If you are graduating this semester and therefore not able to take neural networks in the Fall, you may want to play with some very simple neural networks. You can contrast the linear adaptive filters with simple but nonlinear neural networks.