I can feel it …

I can sense it …
A new for m of int elligence…
is emer ging fr om…
t he ot her side of t he t able !
From Brains to Machines
Dr. Janaka Wijayakulasooriya
Int r oduct ion t o Ar t ificial Int elligence ( AI )
Department of Electrical and Electronic Engineering
University of Peradeniya
Sri Lanka
Our Road Map
§ What is AI ?
§ How it evolved ?
§ What can we do wit h it ?
§ How can we use AI ?
A Chinese Proverb
Tel l me…I wi l l f or get
Show me…I wi l l r emember
I nvol ve me…I wi l l under st and
What is Intelligence ?
Abilit y t o…
§ Think
§ Analyse
§ Lear n
§ Reason
§ Do cr eat ive wor k
§ Per cept ion
§ …
What is Intelligence ?
Int elligence consist s of t he following skills:
§ abilit y t o r eason
§ abilit y t o acquir e and apply knowledge
§ abilit y t o manipulat e and communicat e ideas
Has under st anding/
int ent ionalit y
See
Hear
Touch
Tast e
Smell
Mind
INPUTS
INTERNAL
PROCESSES
OUTPUTS
Senses envir onment
Can Reason
Has knowledge
Exhibit s behaviour
An Intelligent Entity
What is AI ?
§ Media cr eat es t he wor ld t oday
§ AI =
§ 2001: Space odessey
§ St ar Tr ek
§ AI: The movie
§ Oft en por t r ayed as
§ A pr oper t y of an evil comput er
§ Comput er s doing impossible t hings
§ Books and movies
§ Inspir ed many AI r esear cher s
§ Raised t he expect at ions of gener al public
What is AI ?
What others have to say
What is AI ?
§ Aut omat ion of human behaviour
§ Mimic human r easoning pr ocess
§ Knowledge r epr esent at ion in machines
§ Lear ning fr om mist akes ( adapt ive )
…..
“ Is it r eplicat ing human int elligence ?”
What about Emotions ?
What about Emotions
“ I want a voice r ecognit ion soft war e which will make
t he comput er fear s me, when I r aise my voice”
Its 2007, but where is HAL ?
Ar t ificial
int elligence is
about mor e
t han t alking
comput er s and
r obot s in
sear ch of love
and laught er .
In fact , AI is
most useful in
it s simplest
for m
Is it replicating human intelligence ?”
§ Not necessar ily
§ Somet imes mor e…somet imes less
§ It is like
§ Physical Vs Elect r onic Books
§ Act ual Vs Vir t ual shopping
§ Bir d Vs Air planes
… or simply somet hing Vs essence of somet hing
Strong AI Vs Weak AI
§ St r ong AI
§ Comput er s can be made t o t hink in a level at
least equal t o humans
§ Weak AI
§ Adding “ t hinking like” feat ur es t o machines t o
make t hem mor e useful
§ Examples: Exper t syst ems, speech r ecognit ion,
nat ur al language pr ocessing…
Human/Biological Intelligence
§ Thinking humanly ( Cognit ive Modeling ]
§ Cognit ive science
§ 1960s – infor mat ion pr ocessing behavior ism as
t he dominant view in psychology
§ Cognit ive neur oscience
§ Neur ophysiologic basis of int elligence and
behaviour
§ Act ing humanly ( Oper at ional Int elligence ]
§ Tur ing t est
§ Requir ed: Knowledge, r easoning, language
under st anding, lear ning
§ Pr oblem: It is not r epr oducible or amenable t o
mat hemat ical analysis à r at her subject ive
Why should I learn AI ?
Will it get me a Job ?
Goals of AI
Engineer ing Goal:
To solve r eal wor ld pr oblems
To build syst ems t hat
exhibit s int elligent behaviour
Scient ific Goal:
To under st and comput at ional
mechanisms needed for
modeling int elligent behaviour
Goals of AI
Intelligent Systems
§ Many scient ist s believe t hat only t hings t hat can be dir ect ly obser ved
ar e “scient ific”
§ Ther efor e if a machine behaves “as if it wer e int elligent ” it is
meaningless t o ar gue t hat t his is an illusion.
§ Tur ing was of t his opinion and pr oposed t he “Tur ing Test ”
§ This view can be summar ized as:“If it walks like a duck, quacks like a
duck and looks like a duck - it is a duck”
Behaviourist View on Intelligent Machines
A brief history of AI
§ Gest at ion (4 3 -5 6 ):
§ aut omat a t heor y, neur al net wor ks, checker s, t heor em pr oving.
§ Shannon, Tur ing, Von Neumann, Newell and Simon, Minsky,
McCar t hy, Dar mout h Wor kshop.
§ Gr eat expect at ions (5 2 -6 9 ):
§ comput er s can do mor e t han ar it hmet ics!
§ Gener al Pr oblem Solver (GPS), bet t er checker s
§ LISP (LISt Pr ocessing language)
A brief history of AI
A brief history of AI
§ A dose of r ealit y (6 6 -7 4 ):
§ ELIZA: human-like conver sat ion.
§ limit at ions of neur al net wor ks, genet ic algor it hms, machine
evolut ion.
§ act ing in t he r eal wor ld: r obot ics.
§ Knowledge-based syst ems (6 9 -7 9 ):
§ domain focus: exper t s syst ems vs. Gener al Pr oblem Solver s.
§ DENDRAL, MYCIN, XCON, et c.
§ Commer cial AI: t he ‘8 0 s boom (8 0 -9 0 )
§ DEC’s R1 comput er configur at ion pr ogr am
§ many exper t syst ems t ools companies (most ly defunct ): Symbolics,
Teknolwedge, et c.
§ Japan’s 5 t h gener at ion pr oject : PROLOG.
§ limit ed success in aut onomous r obot ics and vision syst ems.
A brief history of AI
Degr ee
of
Mot ivat ion
Dar t mout h
Confer ence
AI
Wint er
New
Technology
Suppor t
Time
1 9 4 8
1 9 7 0 s - 8 0 s
mid-1 9 8 0 s
Japan 5 t h
Gener at ion
Comput er
Test for intelligence: Turing Test
Early Attempts: Eliza
§ Developed in 1 9 6 4 -1 9 6 6 by Joseph
Weizenbaum in MIT
§ Models (par odies) t he r ôle of a Roger ian
psychot her apist engaged in an init ial
int er view wit h a pat ient . Much or t he
t echnique of t he Roger ian psychot her apist
involves dr awing t he pat ient out by r eflect ing
t he pat ient ’s st at ement s back at him.
Do you agree with Turing Test?
John Searle - Chinese Room
§ Paper s ar e passed in wit h
Chinese wr it ing.
§ Human being (who does not
know Chinese) uses book t o
conver t input squiggles t o
out put squiggles
§ Human being copies squiggles
ont o paper s and sends t hem out
of t he r oom.
John Sear le (1932-)
Philosophy pr ofessor
Univer sit y of Califor nia says ...
Outside the Chinese Room
§ Behaviour
§ Chinese quest ions go int o t he r oom.
§ Chinese answer s come out of t he r oom.
§ Funct ionalist int er pr et at ion
§ Whoever is in t he r oom
under st ands Chinese.
Same situation with our Students ?
Do we r eally under st and … or
Just CUT & PASTE
Objections to Turing Test
Objections to Turing Test
Foundations of AI
§ Philosophy: Ar ist ot le, mechanist ic views, mat er ialism,
posit ivism, r at ionalit y.
§ Mat hemat ics: algor it hms, logic, for malizat ion of
mat hemat ics, incomplet eness, decision t heor y.
§ Psychology: behavior ism, cognit ive science.
§ Linguist ics: gr ammar s, synt ax and semant ics.
§ Comput er Science: comput er s, soft war e, t heor y
§ Ot her s: neur oscience, economics, game t heor y.
Ready to play a Game ?
Tic-Tac-Toe
O
O
X
Game of Nim
Another Game ?
§ Robot s
§ Chess-playing pr ogr am
§ Voice r ecognit ion syst em
§ Speech r ecognit ion syst em
§ Gr ammer checker
§ Pat t er n r ecognit ion
§ Medial diagnosis
§ Syst em malfunct ion r ect ifier
§ Game Playing
§ Machine Tr anslat ion
§ Resour ce Scheduling
§ Exper t syst ems (diagnosis,
advisor y, planning, et c)
§ Machine lear ning
§ Int elligent int er faces
A Robot Colony
§ Pr oblem Definit ion:
Intelligent Agents
§ What is an int elligent agent ?
§ St r uct ur e of int elligent agent s
§ Envir onment s
§ Examples
Intelligent agents: their environment and
actions
Ideal rational agents
§ For each possible per cept sequence, an ideal r at ional
agent should t ake t he act ion t hat is expect ed t o maximize
it s per for mance measur e, based on evidence fr om t he
per cept sequence and it s built -in knowledge.
§ Key concept :
mapping fr om per cept ions t o act ions
§ Differ ent ar chit ect ur es t o r ealize t he mapping
Structure of intelligent agents
§ Agent pr ogr am: a pr ogr am t hat implement s t he
mapping fr om per cept s t o act ions
§ Ar chit ect ur e: t he plat for m t o r un t he pr ogr am
(not e: not necessar ily t he har dwar e!)
§ Agent = ar chit ect ur e + pr ogr am
§ Examples:
§ medical diagnosis - par t -picking r obot
§ sat ellit e image analysis - int er act ive t ut or
§ r efiner y cont r oller - flight simulat or
Introduction
§ Humans solve pr oblems by combining
§ Fact s
§ Knowledge
§ They t ake
§ Finding ways t o pr ocess t hem t o r each t he
desir ed solut ion(s)
§ This pr ocess is called as Reasoning
§ Def: Reasoning
§ The pr ocess of wor king wit h knowledge,
fact s and pr oblem solving st r at egies t o
dr aw conclusions
Deductive Reasoning
§ Deduce new infor mat ion fr om logically r elat ed
known infor mat ion
§ Eg. Sher lock Holmes
§ Deduct ive Reasoning uses
§ Axioms (pr oblem fact s)
§ I am st anding in t he r ain
§ Implicat ions
(Relat ed gener al knowledge in t he for m of r ules)
Eg: If I st and in t he r ain THEN I will get wet
§ Conclusion:
We can conclude I will get wet
( ) B B A A ⇒ → ∩
Inductive Reasoning
§ Used t o ar r ive at a gener al conclusion fr om a
limit ed set of fact s
§ Eg:
§ Pr emise: Cr ows in Sr i Lanka can fly
§ Pr emise: Cr ows in India can fly
§ Conclusion: Cr ows can fly
§ For a set of object s, X= { a,b,c,d,...} if pr oper t y P is t r ue for
a, and if P is t r ue for b and if P is t r ue for c, t hen P is t r ue for
all X
§ Not t r ue always
Abductive Reasoning
§ Abduct ion is a for m of deduct ion t hat allows for
plausible infer ence (i.e. conclusions might follow
fr om t he available infor mat ion, but it might be
wr ong)
§ Eg.
§ Gr ound is wet IF it is r aining
§ Gr ound is wet
§ Is it r aining?
Analogical Reasoning
§ For mat ion of models dr awing analogies and
differ ences bet ween t wo object s
§ Eg.
§ Lion is like a t iger
§ Bot h eat meat
§ Live in India
§ However , t hey have differ ent colour , belongs t o
differ ent families
Common-Sense Reasoning
§ Relies mor e on good judgment r at her t han on exact
logic
§ Refer r ed t o as Heur ist ic ; Rule of t humb (best fir st
sear ch)
§ Eg.
§ A loose fan belt usually causes st r ange noises
§ A pat ient wit h TB usually have light fever in evenings
§ How can I dr ive fr om her e t o Kandy ?
§ Valuable in applicat ions t hat r equir es quick
solut ions
Non-Monotonic Reasoning
§ Dur ing t he cour se of solving a pr oblem,
t he st at e of var ious fact s r emains
const ant (Monot onic Reasoning)
§ In non-monot onic r easoning st at e of t he
fact s can change
§ Eg.
§ RULE: If it r ains I will get wet
§ FACT: It is r aining à I will get wet
§ FACT: It ’ s not r aining à I won’ t get wet
§ Non-monot onic r easoning can adjust
Inference
§ Def: The pr ocess used in an exper t syst em t o
der ive solut ions or new infor mat ion fr om known
infor mat ion
§ Infer ence Engine is t he module used for t his
§ It cont ains:
§ What quest ions t o ask t he user ?
§ How t o sear ch t he knowledgebase
§ How t o pick a r ule t o fir e?
§ How t he concluded info. Influence t he sear ch
The world of a wumpus
PIT
PIT
PIT
Stench
Breeze
Prize
Hunter à
Wumpus à
Exploring the world of Wumpus
Knowledge Base:
St ench: No
Br eeze: No
Pr ize: No
OK: [ 1,1]
PIT: - Not in [1,1]
Wu: - Not in [1,1]
Inference:
OK: [2,1] and [1,2]
PIT: Not in [1,2], [2,1]
Wu: Not in [1,2], [2,1]
Action:
à Move t o (1,2)
PIT
PIT
PIT
Exploring the world of Wumpus
Knowledge Base:
St ench: No
Br eeze: Yes
Pr ize: No
OK: [1,2], [2,1]
PIT: - Not in [1,1], [1,2], [2,1]
Wu: - Not in [1,1], [1,2], [2,1]
Inference:
Pit in [3,1] or [2,2] or [1,1]
à Pit in [3,1] or [2,2]
Wu: Not in [1,1],[2,2] and [3,1]
Actions:
§ à Move t o (1,1)
§ à Move t o [1,2]
PIT
PIT
PIT
Exploring the world of Wumpus
Knowledge Base:
St ench: Yes
Br eeze: No
Pr ize: No
OK: [1,2], [2,1]
PIT: Not in [1,1], [1,2], [2,1]
In [2,2] or [3,1] or bot h
Wu: Not in [1,1], [1,2], [2,1], [2,2], [3,3]
Inference:
Wu: in [1,3], [2,2] or [1,1]
è Wu in [1,3]
Pit : not in [1,3], [2,2] or [1,1]
è Pit : In [3,1]
è [2,2] OK
Act ion:
Move t o [2,2]
PIT
PIT
PIT
Exploring the world of Wumpus
Knowledge Base:
St ench: Yes
Br eeze: No
Pr ize: No
OK: [1,2], [2,1]
PIT: Not in [1,1], [1,2], [2,1]
In [2,2] or [3,1] or bot h
Wu: Not in [1,1], [1,2], [2,1], [2,2], [3,3]
Infer ence:
Wu: in [1,3], [2,2] or [1,1]
è Wu in [1,3]
Pit : not in [1,3], [2,2] or [1,1]
è Pit : In [3,1]
PIT
PIT
PIT
Agent
Type
Percepts Actions Goals Environment
Driver
Camera
GPS
Speedometer
Sonar
Video
cameras
Steer
Accelerate
Brake
Safety
Speed
Legal
Profit
Roads
Drivers
Traffic
Pedestrians
Customers
Illustrative example: taxi driver
Table-Driven Agents
function Table-driven-agent(percept) returns action
static: percepts, a sequence, initially empty
table, indexed by percept sequences (given)
append percept to the end of percepts
action := LOOKUP(percepts, table)
return action
• Keeps a list of all percepts seen so far
• Table too large
• takes too long to build
• might not be available
Artificial Neural Networks
Brain and Machine
• The Brain
– Pattern
Recognition
– Association
– Complexity
– Noise Tolerance
•The Machine
–Calculation
–Precision
–Logic
The contrast in architecture
• The Von Neumann architecture
uses a single processing unit;
– Tens of millions of operations per
second
– Absolute arithmetic precision
•The brain uses many slow
unreliable processors acting in
parallel
The Structure of Neurones
axon
cell body
synapse
nucleus
dendrites
axon
cell body
synapse
nucleus
dendrites
The Structure of Neurones
• A neurone only fires if its input signal
exceeds a certain amount (the threshold) in
a short time period.
• Synapses vary in strength
– Good connections allowing a large signal
– Slight connections allow only a weak signal.
– Synapses can be either excitatory or inhibitory.
S
j
f (S
j
)
X
j
a
o
a
1
a
2
a
n
+1
w
j0
w
j1
w
j2
w
jn
A Classic Artifical Neuron
Multilayer Perceptron
Output Values
Input Signals (External Stimuli)
Output Layer
Adjustable
Weights
Input Layer
§ Gener al ar chit ect ur e
Single layer
net input t o Y:
bias b is t r eat ed as t he weight fr om a special unit wit h
const ant out put 1.
t hr eshold r elat ed t o Y
out put
classify int o one of t he t wo classes

·
+ ·
n
i
i i
w x b net
1
¹
'
¹
<

· ·
θ
θ
net
net
net f y
if 1 -
if 1
) (
θ
Y
xn
x1
1
w
n
w
1
b
) ...... , (
1 n
x x
§ Decision r egion/boundar y
n = 2, b != 0, θ = 0
is a line, called decision boundar y, which par t it ions t he plane
int o t wo decision r egions
If a point /pat t er n is in t he posit ive r egion, t hen
, and t he out put is one (belongs t o
class one)
Ot her wise, , out put –1 (belongs t o
class t wo)
n = 2, b = 0, θ != 0 would r esult a similar par t it ion
2
1
2
1
2
2 2 1 1
or 0
w
b
x
w
w
x
w x w x b
− − ·
· + +
2
x
1
x
+
-
) , (
2 1
x x
0
2 2 1 1
≥ + + w x w x b
0
2 2 1 1
< + + w x w x b
Case Studies
PC Based Fet al Moni t or i ng Uni t
Hear t Beat wave shape
Neur al Net wor k
1
2
3
10
1
2
3
15
.
.
.
.
.
.
.
.
W
1,1,1
W
1,1,2
Inputs
(15 at a time)
First layer
1
2
3
10
.
.
.
.
W
2,1,1
W
2,1,2
Second layer
W
3,1,1
W
3,3,1
W
3,2,1
W
3,10,1
Third layer
Out put
Z
-1
Z
-1
Z
-1
Tr aining t he Neur al Net wor k
Tr aining per f or mances
Neur al Net wor k out put
I nput har t beat var iat ion
Recognit ion
User I nt er f ace
4
3
2
1
5
6
Fetal Heart Monitor –Patient No. 03
Base line:110 Print
DECELERATION WARNING !
B.L. Var. :+1 Heart rate:110
Example
Automated character recognition
C:\ocrsamp\janaka7.tif
C:\ocrsamp\janaka11.tif
Example
software to recognize handwritten
characters...
Methodology
Step 1: Taking the letters apart
Methodology
Step 2: Standardizing the letters
1. Skeltonizing
2. Normalizing
Methodology
Step 3: Extracting Features
16 features
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Mean (F1..F8)
M
e
a
n

(
F
9
.
.
F
1
6
)
a
b
c
d
e
f
g
h
i
j
k
l
m
n
o
p q
r
s
t
u
v
w
x
y
z
s
h a
r
a
n
y
a
p e
r
a
d
e
n
i
y
a
a
b
c
d
e
f
g
h
i
j
k
l
m
n
o
p
q
r
s
t
u
v
w
x
y
z
a
b
c
d
e
f
g
h
i
j
k
l
m
n
o
p
q
r
s
t
u
v
w
x
y
z
a
b
c
d
e
f
g
h
i
j
k
l
m
n
o
p
q
r
s
t
q
u
i
c
k
b
r
o
w
n
p
e
r
a
d
e
n
i
y
a
t
e
s
t
i
n
g
w
i
j
a
y
a
k
u
l
a
s
o
o
r
i
y
a
f
a
c
u
l
t
y
o
f
e
n
g
i
n
e
e
r
i n
g
j
a
n
a
k
a
a
n
n
u
a
l
r
e
s
e
a
r
c
h
q
u
i
c
k
b
r
o
w
n
f
o
x
j
u
m
p
s
o
v
e
r
t
h e
l
a
z
y
d
o
g
t
h
a
n
k
y
o
u
f
o
r
c
o
m
i
n
g
t
o
t
h
i
s
e
v
e
n
t
a
b
c
d
e
f
g
h
i
j
k
l
m
n
o
p
q
r s
t
u
v
w
x
y
z
k
u
r
u
n
e g
a
l
a
e
n
g
l
a
n
d
p
a
n
a
d
u
r
a
m
o
n
a
r
a
g
a
l
a
p
e
n
i
d
e
n
i
y
a
k
o
s
h
i
n
n
a
b
a
l
a
n
g
o
d
a
k
u
l
i y
a
p
i
t
i
y
a
h
i
n
d
a
g
a
l
a
g
a
m
p
o
l
a
n
i
l
l
a
m
b
e
p
o
k
u
n
u
w
i
t
a
w
e
l
i
g
a
l
l
a
q
u
l
c
k
b
r
o
w
n
f
o
x
j
u
m
p
s
a
b
c
d
e
f
g
h
i
j
k
l
m
n
o
p
q
r
s
t
u
v
w
x
y
z
q
u
i
c
k
b
r
o
w
n
f
o
x
j
u
m
p
s
o
v
e
r
t
h
e
l
a
z
y
d
o
g
a
b
c
d
e
f g
h
i
j
k
l
m
n
o
p
q
r
s t
u
v
w
x
y z
q
u
i
c
k
b
r
o
w
n
f
o
x
j
u
m
p
s
o
v
e
r
t
h
e
l
a
z y
d
o
g
t
h
i
w
a
n
k
a
w
i
j
e
k
o
o
n
s
e
s
s
i
o
n
a
n
n
u
a
l
r
e
s
e
a
r
c
h
o
o
o
o
o
o
o
o
oooo
oooo
o
o
o
o
o
o
o
o
o
o o
o
o
o
o
o
o
o
o
o
o
a
a
a
a
a
a
a
a
a
a
a
a a
a
a
a a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a a
a
a
a
a
a
a
a
a
a
a
a
a
aa
a
a
a
a
a
u
u
u
u u
u
u
uu
u
u
u
u
u
u
u
u
u
u
u
u u
u
u
u
u e
e
e
e
e
e
e
e
e
e
e
e
e e
ee
ee
e
e
e e
e
e
e
e
e
e
e
e e
e
e
e
e
Feature Space
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Mean (F1..F8)
M
e
a
n

(
F
9
.
.
F
1
6
)
i
i
i
i
i
i
i
i
i
i
i
i
i
ii
i
i
i
i
i
i
i
i
i
i
i
i
i
i
i
i i
i
o
o
o
o
o
o
o
o
oooo
oooo
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
a
a
a
a
a
a
a
a
a
a
a
a a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
aa
a
a
a
a
a
u
u
u
u u
u
u
uu
u
u
u
u
u
u
u
u
u
u
u
u u
u
u
u
u e
e
e
e
e
e
e
e
e
e
e
e
e e
ee
ee
e
e
e e
e
e
e
e
e
e
e
e e
e
e
e
e
Feature Space
Self Adaptive Artificial
Neural Network (SAANN)
Classifier
Char ac t er
Feat ur es
Char ac t er
EXPERI ENCE
SAANN
With only 10 type fonts
With 3 new users
With 1 user
Example
I nt el l i gent Power Qual i t y Moni t or i ng Syst em
Research Background
Voltage and current
waveforms of the
electrical supply is
considered to be
pure sinusoidal
However, practically
these waveforms
deviate from its pure
sinusoidal form due to
many reasons
Research Background
The deviation of voltage or
current waveforms from its
pure sinusoidal form can be
defined as a
Power Quality (PQ)
disturbance
Research Background
During last few decades...
…equipment
sensitive to PQ
disturbances...
have increased dramatically...
…equipment
causing PQ
disturbances
…..The research has developed a novel Intelligent Power
Quality Monitor ( IPQM ), which is capable of…..
… ….PQ di st ur banc es .PQ di st ur banc es
Det ec t i ng Det ec t i ng
Capt ur i ng Capt ur i ng
Cl assi f yi ng Cl assi f yi ng
Loc at i ng Loc at i ng
Anal ysi ng Anal ysi ng
… ….Har moni c sour c es .Har moni c sour c es
Di st r i but ed
Moni t or i ng
Uni t
[ DMU ]
Cent r al i sed
Moni t or i ng
Uni t (CMU)
Vol t age Wavef or ms Vol t age Wavef or ms
Vol t age Wavef or ms
I PQMS
Bus 1
Bus 2
Bus N
Di st r i but ed
Moni t or i ng
Uni t
[ DMU ]
Di st r i but ed
Moni t or i ng
Uni t
[ DMU ]
Capt ur ed
PQ Event s
Capt ur ed
PQ Event s
Capt ur ed
PQ Event s
Di st r i but ed
Moni t or i ng
Uni t
[ DMU ]
Vol t age Wavef or m V(t )
Di st r i but ed Moni t or i ng Uni t
Bus
Di st ur banc e Ext r ac t i on Modul e
Di st ur banc e Ext r ac t i on Modul e (DMU)
Extracts PQ disturbances from the supply voltage signal
0 50 100 150 200 250 300 350 400
-30
-20
-10
0
10
20
30
40
50
0 50 100 150 200 250 300 350 400
-4
-3
-2
-1
0
1
2
3
4
DMU DMU
Ext r ac t ed Di st ur banc e Ext r ac t ed Di st ur banc e
Vol t age Si gnal Vol t age Si gnal
e(t ) e(t )
( )
dt
T
t Vr t v j t Vr t v C
where
r
C
T
t Vr C t Vr C
t v t e
T

]
]
]

− ⋅ ⋅ + ⋅ ·
·
− ⋅ + ⋅
− ·

0
)
4
( ) ( ) ( ) (
,
C

)
4
( ) Im( ) ( Re
) ( ) (
0 0 . 2 0 . 4 0 . 6 0 . 8 1 1 . 2 1 . 4
- 2
- 1
0
1
2
t i m e ( s )
V
o
l
t
a
g
e

(
p
.
u
.
)
0 0 . 2 0 . 4 0 . 6 0 . 8 1 1 . 2 1 . 4
- 1 . 5
- 1
- 0 . 5
0
0 . 5
1
t i m e ( s )
V
o
l
t
a
g
e

(
p
.
u
.
)
Exampl e
0 0. 2 0. 4 0. 6 0. 8 1 1. 2 1. 4
-2
-1
0
1
2
t i me ( s )
V
o
l
t
a
g
e

(
p
.
u
.
)
0 0. 2 0. 4 0. 6 0. 8 1 1. 2 1. 4
- 1. 5
-1
- 0. 5
0
0. 5
1
t i me ( s )
V
o
l
t
a
g
e

(
p
.
u
.
)
V(t)
e(t)
Exampl e
0 0. 2 0. 4 0. 6 0. 8 1 1. 2 1. 4
-2
-1
0
1
2
t i me ( s )
V
o
l
t
a
g
e

(
p
.
u
.
)
0 0. 2 0. 4 0. 6 0. 8 1 1. 2 1. 4
- 1. 5
-1
- 0. 5
0
0. 5
1
t i me ( s )
V
o
l
t
a
g
e

(
p
.
u
.
)
V(t)
e(t)
Exampl e
0 0. 2 0. 4 0 .6 0. 8 1 1 .2 1. 4
-2
-1
0
1
2
t i me ( s )
V
o
l
t
a
g
e

(
p
.
u
.
)
0 0. 2 0. 4 0 .6 0. 8 1 1 .2 1. 4
-0. 3
-0. 2
-0. 1
0
0. 1
0. 2
t i me ( s )
V
o
l
t
a
g
e

(
p
.
u
.
)
V(t)
e(t)
Exampl e
0 0. 2 0. 4 0 .6 0. 8 1 1 .2 1. 4
-2
-1
0
1
2
t i me ( s )
V
o
l
t
a
g
e

(
p
.
u
.
)
0 0. 2 0. 4 0 .6 0. 8 1 1 .2 1. 4
-0. 3
-0. 2
-0. 1
0
0. 1
0. 2
t i me ( s )
V
o
l
t
a
g
e

(
p
.
u
.
)
V(t)
e(t)
Di st r i but ed
Moni t or i ng
Uni t
[ DMU ]
Vol t age Wavef or m V(t )
Di st r i but ed Moni t or i ng Uni t
Bus
Di st ur banc e Ext r ac t i on Modul e
Ext r ac t ed Di st ur banc e
wavef or m e(t )
Event I dent i f i c at i on Modul e
e(t ) Buf f er e
i -1
(T) e
i
(T) e
i +1
(T)
Event I dent i f i c at i on
Modul e ( EI M )
Event I dent i f i c at i on Modul e
Capt ur i ng PQ Di st ur banc es
St eady
St at e
Tr ansi t i on
St at e
I nt er medi at e
St eady St at e I nt er medi at e
Tr ansi t i on St at e
St at e Model
Capt ur ed
Event
Capt ur ed
Event
Capt ur ed
Event
Exampl e
0 0.05 0.1 0.15 0.2
-2
0
2
time (s)
v
o
l
t
a
g
e

(
p
u
)
(a)
0 0.05 0.1 0.15 0.2
-1
0
1
time (s)
v
o
l
t
a
g
e

(
p
u
)
(b)
0 0.05 0.1 0.15 0.2
1
1.5
2
time (s)
S
T
A
T
E
(c)
0 0.005 0.01 0.015 0.02
-1
0
1
time (s)
V
o
l
t
a
g
e

(
p
u
)
(d)
V(t)
e(t)
Extracted
event
wvaeform
Exampl e
0 0.05 0.1 0.15 0.2
-2
0
2
time (s)
v
o
l
t
a
g
e

(
p
u
)
(a)
0 0.05 0.1 0.15 0.2
-1
0
1
time (s)
v
o
l
t
a
g
e

(
p
u
)
(b)
0 0.05 0.1 0.15 0.2
1
1.5
2
time (s)
S
T
A
T
E
(c)
0 0.005 0.01 0.015 0.02
-1
0
1
time (s)
V
o
l
t
a
g
e

(
p
u
)
(d)
V(t)
e(t)
Extracted
event
wvaeform
Di st r i but ed
Moni t or i ng
Uni t
[ DMU ]
Vol t age Wavef or m V(t )
Di st r i but ed Moni t or i ng Uni t
Bus
Di st ur banc e Ext r ac t i on Modul e
Ext r ac t ed Di st ur banc e
wavef or m e(t )
Event I dent i f i c at i on Modul e
Capt ur ed t r ansi t i on event s Capt ur ed st eady-st at e event s
Di st r i but ed
Moni t or i ng
Uni t
[ DMU ]
Cent r al i sed
Moni t or i ng
Uni t (CMU)
Vol t age Wavef or ms Vol t age Wavef or ms
Vol t age Wavef or ms
I PQMS
Bus 1
Bus 2
Bus N
Di st r i but ed
Moni t or i ng
Uni t
[ DMU ]
Di st r i but ed
Moni t or i ng
Uni t
[ DMU ]
Capt ur ed
PQ Event s
Capt ur ed
PQ Event s
Capt ur ed
PQ Event s
Cent r al i sed
Moni t or i ng
Uni t
[ CMU ]
Capt ur ed PQ event s
Cent r al i sed Moni t or i ng Uni t
DMU
Feat ur e Ext r ac t i on Modul e
Feat ur e Ext r ac t i on Modul e (FEM)
Extracting features from PQ Disturbances
Captured
Transition
Event
Waveforms
Steady State Disturbance Feature Vector
( 2 elements )
Transition Disturbance Feature Vector
( 63 elements )
Feature
Extraction
Module (FEM )
Transition Feature
Extractor
( Using DWT )
Steady State Feature
Extractor
( Using FFT )
Captured
Steady State
Event
Waveforms
Exampl e: I mpul si ve Tr ansi ent
Exampl e: Osc i l l at or y t r ansi ent
Ext r ac t ed f eat ur es
Event Cl assi f i c at i on Modul e
Cent r al i sed
Moni t or i ng
Uni t
[ CMU ]
Capt ur ed PQ event s
DMU
Feat ur e Ext r ac t i on Modul e
Cent r al i sed Moni t or i ng Uni t
Event
Classification
Module (ECM)
Transition Event
Classifier
( SAANN-1 )
Steady State Event
Classifier
( SAANN-2 )
Transition Event Feature Vector
( 63 elements )
Steady State Event Feature Vector
( 2 elements )
Oscillatory
Transient
Voltage
Sag
Impulsive
Transient
Voltage
Swell
Momentary
Supply
Interruption
Transition Event Classes
Over
Voltage
Supply
Interruption
Under
Voltage
Steady State Event Classes
Harmonic
Distortion
Normal
Condition
Event Cl assi f i c at i on
Cl assi f yi ng PQ Di st ur banc es
SI MULATI ON
Biometric Recognition System
§ Facial Recognition
§ Voiceprint Recognition
§ Fingerprint Recognition
§ Retinal Patterns Recognition
§ DNA Recognition
ECG signal classification
Thank You !