![[Company Logo Image]](../../images/mlp.bmp) |
Image Processing and Neural Networks Lab
Image Processing and Neural Networks Lab
|
|
Reference Text: "Discrete Random Signals and Statistical Signal Processing" by Charles W. Therrien
This course covers some of the basic algorithms used for the processing of noisy signals in the communications, remote sensing, biomedical imaging, control, energy and defense industries. Students will gain the ability to develop signal models for problems, and devise and implement optimal solutions. Optional problem sessions will be held Friday at 9:00 AM and Saturday at 1:00 PM. We will start by reviewing random processes and classical estimation techniques for the autocorrelation and power spectral density functions. Then we will derive and solve normal equations for optimal filters. The Toeplitz and Levinson-Durbin recursions will be used for (1) deconvolution filter design, (2) Wiener filter design, and (3) AR modeling (which is used in spectral analysis and linear predictive coding (LPC) of speech signals). Adaptive noise cancellation will be introduced. In the last part of the course we will study the maximum likelihood (ML) and maximum a posteriori (MAP) methods of deriving parameter estimation algorithms, along with Cramer-Rao bounds on estimation error variance. The calculation of performance bounds from example data will be discussed. Data compression techniques, including the KLT or principal components transform, will be studied as methods for pre-processing signals for optimal parameter estimators. There will be at least three program assignments in the course. Prerequisites: EE5350 and EE5302, or knowledge of digital filtering and random processes. Autocorrelation Estimator Algorithm Generator Optimal Processor Raw Data Estimator of Statistics and Signal Model Bound Estimator Bounds
Prerequisites: EE4318 or EE5359.054 and EE5302, or knowledge of digital filtering and random processes.
|
Click here to email any
questions or comments about this web site.
|
