Module-1
Discrete Fourier Transforms (DFT): Frequency domain sampling and Reconstruction of Discrete
Time Signals, The Discrete Fourier Transform, DFT as a linear transformation, Properties of the DFT:
Periodicity, Linearity and Symmetry properties, Multiplication of two DFTs and Circular Convolution,
Additional DFT properties.
Text Book:
1. Proakis & Monalakis, “Digital signal processing – Principles Algorithms & Applications”, 4th Edition,
Pearson education, New Delhi, 2007. ISBN: 81-317-1000-9.
2. Oppenheim & Schaffer, “Discrete Time Signal Processing” , PHI, 2003.
3. D. Ganesh Rao and Vineeth P Gejji, “Digital Signal Processing” Cengage India Private Limited, 2017,
ISBN: 9386858231
Consists of three words:
Digital , Signal and Processing
Signal: any (physical or non-physical) quantity that
varies with time, space, or other independent
variable(s)
Digital: a discrete-time and discrete-valued signal, i.e.
digitization involves both sampling and quantization
Processing: operations on the signal
Signals
Continuous-time Discrete-time
Continuous-value Continuous-value Discrete-value
Analog Discrete Digital
4
Signals are everywhere and may reflect countless
measurements of some physical quantity such as:
◦ electric voltages
◦ brain signals
◦ heart rates
◦ temperatures
◦ image luminance
◦ investment prices
◦ vehicle speeds
◦ seismic activity
◦ human speech
Various apparatus could be used to acquire signals,
including:
◦ Digital camera → Image
◦ MRI scanner → Activity of the brain
◦ EEG/EMG/EOG electrodes → Physiological signals
◦ Voice recorder → Audio signal
1D (e.g. dependent on time)
2D (e.g. images dependent
on two coordinates in a plane)
3D (e.g. describing an object in space)
In some applications, signals are generated by multiple
sources or multiple sensors→ represented by a vector
Such a vector is called a multi-channel signal.
Example: brain signals 50
C3 0
50
0 500 1000 1500 2000 2500
50
C4 0
50
0 500 1000 1500 2000 2500
50
P3 0
50
0 500 1000 1500 2000 2500
20
P4 0
20
0 500 1000 1500 2000 2500
20
O1 0
20
0 500 1000 1500 2000 2500
20
O2 0
20
0 500 1000 1500 2000 2500
Sample (250 per second)
Continuous-time signals are signals defined at each
value of independent variable(s).
They have values in a continuous interval (a,b) that
could extend from -∞ to ∞.
Discrete-time signals are defined only at specific values
of independent variable(s).
Discrete-time signals are represented mathematically
by a sequence of real or complex numbers.
CT DT
Continuous function V of Discrete function Vk of discrete
continuous variable t (time, sampling variable tk, with k =
space etc) : V(t). integer: Vk = V(tk).
0.3 0.3
0.2 0.2
Voltage [V]
Voltage [V]
0.1 0.1
0 0
-0.1 -0.1 ts
-0.2 -0.2
0 2 4 6 8 10 0 2 4 6 8 10
time [ms] sampling time, tk [ms]
Periodic sampling
Both continuous and discrete-time signals can take a finite
(discrete) or infinite (continuous) range.
For a signal to be called digital, it must be discrete-time and
discrete-range, i.e. digitization involves both sampling and
quantization.
Signals could be deterministic, with an explicit
mathematical description, a table or a well-defined rule.
All past, present, and future signal values are precisely
known with no uncertainty:
s1(t) =at S2(x,y)=ax+bxy+cy2
In contrast, for random signals the functional relationship is
unknown.
→ statistical analysis techniques
A system that performs some kind of task on a signal
which depends on the application, e.g.
◦ Communications: modulation/demodulation, multiplexing/de-
multiplexing, data compression
◦ Speech Recognition: speech to text transformation
◦ Security: signal encryption/decryption
◦ Filtering: signal denoising/noise reduction
◦ Enhancement: audio signal processing, equalization
◦ Data manipulation: watermarking, reconstruction, feature
extraction
◦ Signal generation: music synthesis
Digital Signal Processing
Advantages Limitations
• More flexible • A/D & signal processors’ speed
• Data easily stored • Finite word-length effect:
• Better control over accuracy (round-off: Error caused by
requirements rounding math calculation result to
nearest quantization level )
• Reproducibility
• Cheaper
Theoretical vs. Applied
Applicable to any field Easier to comprehend
Algorithm development vs. implementation
e.g., C++-code, e.g., ASIC, DSP chip
Matlab code
Much faster
Easier to adapt
Applications include speech generation / speech recognition
Speech recognition: DSP generally approaches the problem of
voice recognition in two steps: feature extraction followed by
feature matching.
Source: Canon
A common method of obtaining information about a
remote object is to bounce a wave off of it.
Applications include radar and sonar.
DSP can be used for filtering and compressing the data.
Source:
WHIO
Source:
CCTT.org
Pattern recognition is a research area that is closely
related to digital signal processing.
Definition: “the act of taking in raw data and taking an
action based on the category of the data”.
Pattern recognition
classifies data based on
either a priori knowledge
or on statistical information
extracted from the patterns.
Source:
merl.com
The “Biometrics” field
focuses on methods for
uniquely identifying
humans using one or more
of their intrinsic physical
or behavioural traits.
Examples include using
face, voice, fingerprints,
iris, handwriting or the
method of walking.
Source: BBC
• A means for communication
between a brain and a
computer via measurements
associated with brain
activity.
• No muscle motion is
involved (e.g., eye
movement).
50
C3 0
50
0 500 1000 1500 2000 2500
50
C4 0
50
0 500 1000 1500 2000 2500
50
P3 0
50
0 500 1000 1500 2000 2500
20
P4 0
20
0 500 1000 1500 2000 2500
20
O1 0
20
0 500 1000 1500 2000 2500
20
O2 0
20
0 500 1000 1500 2000 2500
Sample (250 per second)
BCI Application- Neuroprosthesis
Hold cup for drinking
http://www.dpmi.tugraz.at/
Reviewed the course outline
Reviewed basic concepts and terminologies of DSP
Examined some practical examples
25
26
27
28
29
30
31
32
46
47
Find the IDFT of 4 point sequence, X(k)=( 4, j2, 0, -j2) using the DFT.
50
51
52
53
54
55
56
