AG

C
DS
P

Digital Signal Processing &

Digital Filters
An Introductory Course
By
Professor A G Constantinides
MSc, EE4, ISE4, PhD

Professor A G
Constantinides© 1
AG
Digital Signal Processing &
C
DS
Digital Filters
P

Contents
1-Introduction
1) Introduction to Digital Signal Processing
 Review of background DSP
 Review of mathematical methods
 Review of discrete-time random
processes and linear systems

Professor A G
Constantinides© 2
AG
Digital Signal Processing &
C
DS
Digital Filters
P
2) Multirate techniques and wavelets
 Introduction to short-time Fourier analysis

 Filter-banks and overlap-add methods of analysis

and synthesis
 Introduction to generalised time-frequency

representation
 Wavelet analysis

 Multirate signal processing

 Interpolation and decimation

 Efficient filter structures for interpolation and

decimation
Professor A G
Constantinides© 3
AG
Digital Signal Processing &
C
DS
Digital Filters
P
3) Classical spectrum estimation methods
 Power spectrum, power spectral density functions,

random processes and linear systems

 Introduction to statistical estimation and estimators

 Biased and unbiased estimators

 Einstein/Wiener Khintchine Theorem

 Estimation of autocorrelations

 Means and variances of periodograms

 Smoothed spectral estimates, leakage

Professor A G
Constantinides© 4
AG
Digital Signal Processing &
C
DS
Digital Filters
P
4) Modern spectrum estimation methods
 Introduction to modern spectral estimation:

Principles and approaches

 Cramer-Rao Lower Bound (CRLB) and Efficient

estimators
 The Maximum Entropy Method (MEM) or

Autoregressive Power Spectrum Estimation:

Principles.
 The MEM equations and Levinson/Durbin

algorithm
Professor A G
Constantinides© 5
AG
Digital Signal Processing &
C
DS
Digital Filters
P

4) Modern spectrum estimation methods

(continued)
 Introduction to Linear Prediction

 Linear Predictive Coding using covariances

and correlations
 Cholesky decomposition

 Lattice Filters

 Linear Prediction of Speech Signals

Professor A G
Constantinides© 6
AG
Digital Signal Processing &
C
DS
Digital Filters
P

5) Adaptive signal processing

 Introduction to adaptive signal processing

 Objective measures of goodness

 Least squares and consequences

 Steepest descent

 The LMS and RLS algorithms

 Kalman Filters

Professor A G
Constantinides© 7
AG
Digital Signal Processing &
C
DS
Digital Filters
P

6) Applications
 Communications

 Biomedical

 Seismic

 Audio/Music

Professor A G
Constantinides© 8
AG
C
DS
DIGITAL FILTERS
P
Digital Filters
 In this course you will learn:

 How to choose an appropriate filter

response.
 Why Butterworth responses are

maximally flat.
 Why Chebyshev and Elliptic responses

are equiripple.
 When to choose an IIR and when an FIR

filter Professor A G
Constantinides© 9
AG
C
DS
DIGITAL FILTERS
P 
How do you design FIR and IIR filters from
specifications on amplitude performance?
 What are multirate systems and their
properties? What is interpolation /
Upsampling and Decimation /
Downsampling?
 How do you design efficient Decimation
and Interpolation systems?
 What are frequency transformations and
how do you design these?
 How accurate is the DFT as a spectrum
estimator? Professor A G
Constantinides© 10
AG
C
DS
DIGITAL FILTERS
P
 What are short FFT algorithms?
 How do you choose the required
wordlength?
 What are Fast Convolutions and
how are they realised?
 How do you deal with a DSP
problem in practice?

Professor A G
Constantinides© 11
 Course content
AG
C
DS
P
Assumed DSP background

DSP Background folder


1-Introduction

2-z transform

3-transfer functions

4-Signal Flow Graphs

5-digital filters intro

Professor A G
Constantinides© 12
 Course content
AG
C
DS
P
2-Digital Filter Design
 1-Digital Filters (FIR)

 2-Digital Filters (IIR)

3-Multirate
1-Interpolation_Decimation

Professor A G
Constantinides© 13
 Course content
AG
C
DS
P
4-Tranforms
 1-DFT

 2-DFT_one2two

 3-general transforms

 4-Wavelets

5-Finite Wordlength
 1-Finite Wordlength
Professor A G
Constantinides© 14
 Course content
AG
C
DS
P
6-Spectrum Estimation (Assumed
background in Mathematical
Background folder)
 1-Fourier transform & DFT
 2-FFT-based Power Spectrum Estimation
 3-Modern Spectrum Estimation
 4-Intro-Estimation
 5-Eigen-based methods
 6-A Prediction Problem
Professor A G
Constantinides© 15
 Course content
AG
C
DS
P
7-Adaptive Signal Processing
 1-Adaptive Signal Processing

8-Applications
 1-Applications

 2-Applications

Professor A G
Constantinides© 16
 Digital Signal Processing &
AG
C Digital Filters
DS
P
BOOKS
 Main Course text books: Digital

Signal Processing: A computer Based

Approach, S K Mitra, McGraw Hill
 Mathematical Methods and Algorithms

for Signal Processing, Todd Moon,

Addison Wesley
 Other books:

 Digital Signal Processing, Roberts &

Mullis, Addison Wesley

 Digital Filters, Antoniou, McGraw Hill

Professor A G
Constantinides© 17
AG
C
DS
DIGITAL FILTERS
P
Analogue Vs Digital Signal Processing
Reliability:
Analogue system performance degrades
due to:
 Long term drift (ageing)
 Short term drift (temperature?)
 Sensitivity to voltage instability.
 Batch-to-Batch component variation.
 High discrete component count
Interconnection Constantinides©
failures
Professor A G
18
AG
C
DS
DIGITAL FILTERS
P
Digital Systems:
 No short or long term drifts.
 Relative immunity to minor power
supply variations.
 Virtually identical components.
 IC’s have > 15 year lifetime
 Development costs
 System changes at
design/development stage only
software changes.
 Digital systemConstantinides©
simulation
Professor A G
is realistic. 19
AG
C
DS
DIGITAL FILTERS
P
Power aspects
 Size

 Dissipation

 DSP chips available as well as

ASIC/FPGA realisations

Professor A G
Constantinides© 20
AG
C
DS
Applications
P
Radar systems & Sonar systems

 Doppler filters.
 Clutter Suppression.
 Matched filters.
 Target tracking.
 Identification
Professor A G
Constantinides© 21
AG
C
DS
DIGITAL FILTERS
P

Image Processing
 Image data compression.

 Image filtering.

 Image enhancement.

 Spectral Analysis.

 Scene Analysis / Pattern

recognition. Professor A G
Constantinides© 22
AG
C
DS
DIGITAL FILTERS
P

Biomedical Signal Analysis

 Spatial image enhancement. (X-

rays)
 Spectral Analysis.

 3-D reconstruction from projections.

 Digital filtering and Data

compression.
Professor A G
Constantinides© 23
AG
C
DS
DIGITAL FILTERS
P

Music
 Music recording.

 Multi-track “mixing”.

 CD and DAT.

 Filtering / Synthesis / Special

effects.
Professor A G
Constantinides© 24
AG
C
DS
DIGITAL FILTERS
P

Seismic Signal Analysis

 Bandpass Filtering for S/N

improvement.
 Predictive deconvolution to extract
reverberation characteristics.
 Optimal filtering. (Wiener and
Kalman.) Professor A G
Constantinides© 25
AG
C
DS
DIGITAL FILTERS
P
Telecommunications and Consumer
Products
These are the largest and most
pervasive applications of DSP
and Digital Filtering
 Mobile Communications

 Digital Recording

 Digital Cameras

 Blue Tooth or similar

Professor A G
Constantinides© 26