Attention and Pattern Perception

parathyroidsanchovyΤεχνίτη Νοημοσύνη και Ρομποτική

17 Νοε 2013 (πριν από 3 χρόνια και 4 μήνες)

65 εμφανίσεις

Attention and Pattern Perception

Chapter 4

September 25, 2012

2

Eight facts about perception

(1
-

4)

1.
Is limited:

Frequency spectrum
,
Infrared Scene
,
Mite
,
Powers of Ten
,
Edgerton Bullet
.

2.
Is selective: cocktail party phenomenon

3.
Requires memory: Recognition

4.
Is not entirely veridical (trustworthy):
Hermann Grid
,
Spiral Illusion
,
Simultaneous Contrast
,
optical

From J.
Pomerantz

Lecture

3

Eight facts about perception

(5
-

8)

5.
Takes time: flicker,
Metacontrast

6.
Corresponds more to the distal than to the proximal
stimulus:
Adelson Shadow Effect

7.
Involves the active organization of sensory information:
Hexagram of Spots
,
R. C. James Photograph

,
Necker
Cube
,

Subjective Necker Cube

8.
Focuses on change, not on steady
-
state information:
disappearance of
stabilized images

From J.
Pomerantz

Lecture

4

And finally…


The key function of perception,
including all its components is survival.


The key information source comes from
the constancies.


Invariants tell you about the world

5

Perception research


Seeks to understand these phenomenon


Perception is essentially needed for determining


Where


What


These processes involve


Bottom up processing


Top down processing

6

Top
-
down vs. Bottom
-
up


Who can tell me what each of these
mean?


How is this distinction important when
considering human perception?

7

What vs. Where
?


Two separate systems in the perceptual
system


Dorsal identifies


Where (or How)


Ventral identifies


What

7

8

Where?


Bottom up


Visual system


Accommodation


changing of the pupil size to focus


Convergence


movement of the two eyes together to focus


Stereopsis


the perception of the depth with binocular vision


Auditory system


Interaural

intensity differences


Interaural

time differences


Top down


Contextual
cues (e.g., framing)


Previous knowledge

9

What?


Global vs. Local features


Object centered perspective


Biederman

geons


Template matching


Prototye



Viewer
centered perspective


Feature
-
matching (global vs. local)


Rotation issues


Gestalt
principles of
grouping

9

Recognition by Components Theory


Recognition by Components Theory


Biederman (1987)


Describes the pattern recognition process in terms of
how people recognize 3
-
D objects

by identifying
basic features that comprise the objects.


These basic elements are composed of an alphabet of
36 primitive shapes, called geons.


Geons


geometric ions


The basic building blocks for identifying 3
-
D objects.

Recognition by Components Theory


These 3
-
D shapes are critical to pattern
recognition because the objects we encounter
in life can be rotated in three dimensions and
create an unlimited number of impressions on
the retina.

Recognition by Components Theory

Geons

are the basic building blocks for identifying three
-
dimensional

objects. Some of them are numbered and shown
on the left.
You can
see how these
geons

compose common objects by matching the
numbers on the left with numbers on the objects on the right.

Recognition by Components Theory

Geon Flashlight.
According to the recognition by components theory
(
Biederman
, 1987), people identify an object by noticing its edges and the

corresponding
geons

that fit the edges and then rely on long
-
term memory
of objects that have that configuration of
geons
. (A) This is a figure composed
of
geons
.
Geons

are each of the small parts (cylinder,
etc
). (B) The object is

deliberately obscured by visual noise but you can still identify its edges, and
therefore, what the object is. (C) The edges are difficult to identify so is it
difficult to find the flashlight in the picture.

A

B

C

Template
-
Matching Theory


Template
-
Matching Theory


We have stored away an unlimited number of
patterns, literal copies corresponding to every
object that we have experienced.


These patterns are labeled with the name of the
object.


Whenever we see a new instance of one of these
objects, it is matched to a stored template that is
instantly activated and informs us of the name of
the pattern we have just experienced.

Template
-
Matching Theory


It is highly impractical, however, when the set
of possible patterns is very large.


In addition to its inefficiency, a template
-
matching system fails to account for our ability
to recognize new objects.


How, for example, do we manage to read the
handwriting of a person we just met?

Prototype Theory


The limitations of template theory can be
dealt with by imagining that the template is
not a literal match with an object, but is an
average or typical instance of the many
different views of that object.


This average is called a prototype and the
theory that uses them for pattern recognition
is called prototype theory.

Prototype Theory


According to the prototype view, pattern
recognition occurs when the features of the
object to be recognized overlap in some way
with the features of the prototype.


Using prototypes to recognize patterns does
not require an exact match between the
object and the prototype, nor does it require
the storage of patterns for every possible view
of an object.

Prototype Theory

Students learned to label dot patterns that were distortions from the
original prototype. They were later able to correctly label new distorted

patterns, but they were especially effective in labeling the original

prototype that they had not previously seen.

Prototype Theory


Ecological validity


Experiments based on how people operate in the
real world are said to have ecological validity.


Many of the later prototype studies have used
stimuli that are more similar to situations in which
we might normally find ourselves.

Prototype Theory

Solso

and McCarthy (1981) created prototypical faces from a police Identikit.
For each prototype, they created exemplars that varied in their degree of
similarity to the prototype (from 75% down to 0% similar). In this learning
phase, the participants had not seen the prototypes. Later, they had to judge
whether new and some old faces had been seen before.

Prototype Theory

The data were graphed, showing that participants rated the prototypes as
old (even though they had never seen them before), and were more
confident of that judgment than all others. The old and new exemplars were
rated as old based on their similarity to the prototype.

Distinctive Features Theory


Global and Local Features


How does the mind reduce the number of
possibilities so that it can concentrate on one
thing?


A partial answer is that we start with global
features and progress to more local ones.

Distinctive Features Theory

In Navon figures, large letters are composed of small letters. When
participants heard the letter name H, they responded “yes” equally
quickly, no matter what the small letters were

these low
-
level features
did not affect the global judgment. However, when deciding the name of
the small letters, they were affected by the name of the capital letter

the global features are processed first.

24

Gestalt Principles of Grouping


Organizing a sea of input


Gestalt principle of grouping


Sum of the parts is
different

than the whole.

25

Similarity

26

Proximity

27

Continuity

28

Closure

29

Grouping


These effects can be very powerful


Raise your hand when you notice the
different item

30

31

Now we

ll add a context

32

33

Raise your hand when you know which one
is different

34

35

Now let

s look at a different effect



Raise your hand when you notice the
different item


36

37

Now we

ll add a context

38

39

Again, raise your hand when you know
which one is different

40

41

What?


Methods of identification


Global vs. Local features


Object centered perspective


Viewer
centered perspective


Gestalt
principles of
grouping


Which of these are top
-
down and which are
bottom up?

41

Face recognition


Do we have a separate ability for recognizing
faces?


Or is it LOTS of practice?


Prosopagnosia


Evidence for special area (double
disassociation studies)


Greebles


Similar to faces and each other


Have same number of parts in the same
configuration


Normals

and patients treat them like objects


Unless they become “
Greeble

experts”


They process
Greebles

similar
to the way
we

process faces.
Greeble

experts also activate
their "fusiform face area" in the brain more than
novices do

42

43

Cognitive Psychology?


The term

cognitive psychology


covers a
lot of ground


Usually associated with so
-
called

higher
-
level


mental processes (reasoning,
memory)


Why is perception part of
cognitive
psychology?



What about applications?