Purdue University
School of Electrical and Computer Engineering

EE648 Wavelet, Time-Frequency, and Multirate Signal Processing

Spring 2003
Prof. Ilya Pollak

Note: Many documents on this website are in Acrobat format. To download a free version of the Acrobat viewer, go to the Adobe web site.


Course TA: Yan Huang, MSEE 176, 494-3465, yanh@ecn.purdue.edu. Office hours: Thursdays 2-3pm.

Course Information.
Tentative syllabus.
Lecture notes for the first few lectures.
Announcements.

Problem Set 1.

Problem Set 2.

Problem Set 3.

Problem Set 4.

Required textbook: S.G. Mallat. "A Wavelet Tour of Signal Processing." 2nd Edition. Academic Press, 1999. ISBN 0-12-466606-X


Other strongly recommended books:

G. Strang, T. Nguyen. "Wavelets and Filter Banks". Wellesley-Cambridge Press, 1997. ISBN 0-9614088-7-1

I. Daubechies. "Ten Lectures on Wavelets". Philadelphia: Society for Industrial and Applied Mathematics, 1992. ISBN 0-89871-274-2

M. Vetterli, J. Kovacevic. "Wavelets and Subband Coding." Prentice Hall, 1995. ISBN 0-13-097080-8


The main theme of this course is sparse representations of signals. Orthogonal bases are of particular interest because they can efficiently approximate certain types of signals with just a few vectors. Two examples of such applications which will be studied in this course are image compression and estimation of noisy signals.

We will consider several approximation schemes. For example, a linear approximation projects the signal over M vectors chosen a priori. Better approximations are obtained by choosing the M basis vectors adaptively, depending on the signal. A further degree of freedom is introduced by choosing the basis itself adaptively, from a family of bases. In this context, a number of very successful recent algorithms for nonlinear approximation, noise removal, and feature extraction will be considered, such as wavelet thresholding, best basis search, and matching pursuit.

Several types of orthogonal bases will be studied: wavelet bases, wavelet packet bases, and local cosine bases. We will study their properties, as well as fast algorithms for calculating the coefficients of a signal with respect to a basis and for reconstructing the signal from its coefficients. We will also address time-frequency properties of various signal decompositions--which is an important tool in deciding what decomposition is appropriate for a particular application.

The pre-requisite for this course is thorough understanding of fundamentals of signal processing--such as Fourier analysis and sampling. These topics will be reviewed at the beginning of the course.

The administration of the course will be very simple: the final grade will be determined solely on the basis of four problem sets which you can think of as take-home exams. Tentative due dates for the problem sets are: February 17, March 10, April 14, April 28. Each problem set will be posted on the course web site approximately one month before its due date. Each problem set is due in class for on-campus students and must be postmarked by the due date for off-campus students. No collaboration with anyone is allowed on the problem sets; however, you can use any literature you wish. There will be no in-class exams and no final.