# Unit information: Statistical Signal and Image Processing in 2019/20

Please note: Due to alternative arrangements for teaching and assessment in place from 18 March 2020 to mitigate against the restrictions in place due to COVID-19, information shown for 2019/20 may not always be accurate.

Please note: you are viewing unit and programme information for a past academic year. Please see the current academic year for up to date information.

Unit name Statistical Signal and Image Processing EENGM0016 10 M/7 Teaching Block 2 (weeks 13 - 24) Professor. Achim Not open EENGM1400 None Department of Electrical & Electronic Engineering Faculty of Engineering

## Description

The aim of this module is to bridge the gap between classical and modern methods of signal analysis, with an emphasis on the processing of stochastic signals. Students will gain an awareness of optimum signal processing methods primarily based on the least squares error criterion. Optimum filter design will be presented, based on the Wiener filter followed by LMS and RLS filter realizations. Three key application uses of this technology are spectrum estimation, noise cancellation and beamforming. These will be covered from a theoretical and application perspective. An introduction to advanced parameter estimation techniques will also be presented.
Elements:

• Introduction; signal classes, stochastic processes, scope, tools, application areas
• Matrix methods and Numerical Linear Algebra; definitions, operations, solution of equations
• Stochastic processes and Parameter Estimation; stationarity, statistics, distributions, orthogonality, introduction to parameter estimation (MMSE, ML, MAP)
• Optimum Least-Squares Filtering; normal equations, Wiener filtering, AR, MA, ARMA models, linear prediction, lattice filters, Kalman filters
• Adaptive Digital Filters; structure and configurations, performance criteria, error surface searching, gradient methods, LMS, RLS, convergence, fast algorithms and numerical stability
• Model Selection and Parameter Estimation; least squares, Bayesian, maximum likelihood
• Spectral Estimation; DFT, FFT (periodogram), model based techniques (MA, AR, ARMA)
• Adaptive Noise Cancelling; reference signal, output SNR, leakage, applications
• Adaptive Beamforming; conventional methods, adaptive methods, constrained and unconstrained methods, direction of arrival estimation

## Intended learning outcomes

On completing this unit, the student will be able to:
1. Quantify the limitations of conventional methods of spectrum estimation.
2. Implement superior model-based algorithms.
3. Design and realise optimum code adaptive digital filters for a range of application scenarios including noise cancellation, linear prediction and beamforming.

Lectures

## Assessment Details

Exam, 2 hours, 100% (All ILOs)