ENEE 630: ADVANCED DIGITAL SIGNAL PROCESSING

Course Goals:

This is the first-year graduate course in signal processing. The objective is to establish fundamental concepts of signal processing on multirate processing, parametric modeling, linear prediction theory, modern spectral estimation, and high-resolution techniques.

Course Prerequisite(s):

ENEE 425 or equivalent; co-requisite ENEE 620 or equivalent.

Reference(s):

• P. P. Vaidyanathan, Multirate Systems and Filter Banks, Prentice-Hall, 1993.
• D.G. Manolakis, V.K. Ingle, and S.M. Kogon, Statistical and Adaptive Signal Processing, Mc- Graw Hill, 2000.
• M. Hayes, Statistical Digital Signal Processing and Modeling, Wiley, 1996.
• S. Haykin, Adaptive Filter Theory, 2nd Ed, Prentice-Hall, 1991.

Core Topics:

1. Multirate Signal Processing (Chaps. 2, 4, 5, Vaidyanathan)
• Decimation and interpolation; sampling rate conversion; direct-form and polyphase representation.
• Time-varying filter structures; implementation of DFT Filter Banks; multistage implementation of sampling-rate conversion
• Quadrature mirror filter (QMF) bank; M-channel filter banks; multiresolution filter banks.
• Perfect reconstruction systems; alias-free filter banks.
• Multi-resolution analysis.
2. Parametric Signal Modeling and Linear Prediction Theory (Chaps. 2, 5, 6, Haykin; or chaps. 4, 6, 9, Manolakis et al)
• Stochastic time-series models: AR, MA, ARMA; Wold decomposition theorem.
• Discrete Wiener filters: principle of orthogonality, normal equations.
• Linear prediction theory: forward and backward linear predictions and their properties.
• Levinson-Durbin algorithm; lattice prediction filter; innovation process; joint-process estimation.
3. Spectral Estimation (Chaps. 5, 9, Manolakis et al)
• Nonparametric methods: Periodograms and windowing methods; statistical properties; minimumvariance spectral estimation.
• Parametric methods: AR, MA, and ARMA spectral estimation; maximum entropy method.
• Higher-order statistics: parametric and non-parametric approaches.
• High-resolution techniques: MUSIC algorithm.