Class Notes (835,638)
Canada (509,305)
Psychology (5,217)
PSYCH 2H03 (134)
Lecture

Ch. 3 - Perception.pdf

6 Pages
117 Views
Unlock Document

Department
Psychology
Course
PSYCH 2H03
Professor
Judith Shedden
Semester
Fall

Description
PNB  2XA3   Chapter  3:  Perception     Beyond  the  info  given   • Perception   ≠  Sensation   • Our  perception  goes  FAR  beyond  the  info  given .  We  see  the  world  in  terms  of   objects.  However,  the  visual  stimulus   (colours,  lines)  does  not  consist  of  objects.  A  lot  of  processing  needs   to  be  done,  and  assumptions  need  to  be  made  by   our  perceptual  system.   • We  see  what  we  expect  to  see.  We  hear  what  we  expect  to  hear.     Pattern  Recognition   • A  pattern  is  a  complex  configuration  of  sensory  signals   • How  do  we  recognize  patterns?   o We’re  easily  able  to  recognize  an  object  even  though  it’s  in  an  orientation  we’ ve  never  seen  before.   o At  early  stages,  it  is  effortless  and  automatic.  We’re  usually  not  even  aware  of  the  basic  sensory  signal,  but  only   of  the  meaning.   o Cells  first  respond  to  line  segments  and  colours  in  V1.  Upper  levels  recognize  objects  and  give  us  the  conscious   recognition  of  what  the  objects  are.  The  upper  levels  are  also  feeding  back  to  V1  so  that  you  can  control  what   part  of  the  visual  field  you  are  looking  at   –  which  lines  or  colours  are  being  focused  on  to  be  processed.   o Which  comes  first?  Global  to  local  recognition  (recognize  the  whole,  followed  by  the  parts )  OR  local  to  global   recognition  (identify  the  parts,  then  construct  the  whole )?   • Bottom-­‐up  Processing:  Data-­‐driven  processing.  Physical  characteristics  of  stimulus  influence  perception.   • Top-­‐down  Processing:  Conceptually-­‐driven  processing.  Knowledge  and  expectations  influence  perception.   o Parsing:  To  disassemble  a  stimulant  into  its  individual  parts     Perceptual  Constancy  (Perception   ≠  Sensation)   • Our  ability  to  perceive  an  object  as  unchanging  even  though  the  visual  image  is  constantly  changing.   • Size  constancy:  Size  of  objects  around  us  are  unchanging.   o Someone  walking  away  does  not  actually  get  smaller.     o Variants  of  the  Muller/Lyer  Illusion:  All  horizontal  lines  are  the  same  width,  but  because  of  the  context,  we   perceive  them  as  different  lengths   • Shape  constancy:  Objects  are  perceived  as  the  same  shape  when  viewed  from  different  angles,  despite  different  retinal   images  of  the  same  object  (i.e.  When  a  door  opens,  we  know  that  the  door  stays  the  same  shape,  even  though  it  l ooks   like  it  is  changing  shapes).   • Location  constancy:  We  perceive  objects  around  us  as  stationary  (i.e.  passing  other  cars  on  he  road). • Brightness  constancy:  Brightness  of  an  object  is  perceived  relative  to  its  background.     • Colour  constancy:  Colour  of  an  object  is  perceived  relative  to  its  background     Gestalt  Principles   • Multistability  (Figure-­‐ground):  The  ability  to  determine  what  aspect  of  a  visual  scene  is  part  of  the  object  itself  and   what’s  part  of  the  background.   • Emergence   • Reification:  The  constructive/generative  aspect  of  conception.  The  perception  that  you  experience  contains  more  spatial   info  than  the  sensory  stimulus  that  you  are  looking  at.  You  are  imposing  more  info  to  what  you  are  looking  at.  We   impose  order  on  nothingness  based  on  arrangement.   • Invariance:  Angles  of  objects  don’t  matter.  We  know  that  the  object  is  still  the  same  object.     Gestalt  Grouping  Processes   • Closure:  Tendency  to  fill  in  gaps  and  perceive  a  whole  object  (imposing  lines).   • Similarity:  Tendency  to  group  together  elements  that  are  physically  similar    • Proximity:  Elements  close  together  in  space  tend  to  belong  together.   • Symmetry   • Continuity:  Tendency  to  perceive  a  simple,  continuous  form  rather  than  a  combination  of  awkward  forms  (i.e.  “X”)   • Common  Fate:  The  idea  that  change  in  the  same  way  should  be  grouped  together  (i.e.  school  of  fish).         Depth  Perception   • Relative  Size:  If  2  objects  are  assumed  to  be  the  same  size,  the  one  producing  a  larger  retinal  image  is  assumed  to  be   closer   • Height  in  the  visual  field :  More  distant  objects  are  higher  in  the  visual  field  than  closer  objects   • Interposition:  Closer  objects  block  the  view  of  objects  further  away   • Linear  perspective:  The  closer  together  two  converging  lines  are,  the  greater  the  distance  is  perceived   • Reduced  clarity:  Greater  distances  produce  less  clarity   • Shading:  Light  and  shadow  contribute  to  the  perception  of  dimensions   • Textural  gradients:  Texture  appears  smaller  and  less  detailed  as  distance  increases   • Structure  through  motion:  Motion  provides  cues  to  3D  properties  of  objects   • Movement  gradient/Motion  parallax :  The  difference  in  relative  movement  provides  cues  to  the   difference  in  distance     Object  Recognition   • Once  the  form  has  been  organized  by  our  perceptual  system  (i.e.  what  parts  belong  together,  what  is  figure  and  what  is   ground),  we  then  need  to  identify  the  object .   • Object  recognition  ≠  object  perception   • Contextual  effects  and  constraint  satisfaction   o Ambiguous  at  the  feature  and  letter  levels   o Disambiguated  at  the  word  level     What  happens  when  the  pattern  recognition  system  is  damaged?   • Prosopagnosia:  Unable  to  recognize  faces.  They  can  tell  it’s  a  face  (including  gender,  age,  etc.),  but  can’t  tell  you  whom   the  face  belongs  to  –  not  even  family  members.   • Object/Visual  agnosia:  Unable  to  recognize  objects  ( i.e.  The  man  who  mistook  his  wife  for  a  hat)     Theoretical  Approaches  to  Object  Recognition   • Template  matching,  feature  analysis,  structural  approach  (recognition  by  components)   • All  explain  some  of  the  data  –  so  our  model  of  recognition  needs  to  be  very  complex,  to  include  many  different  ways,  or   components  of  object  recognition.     Template  Theory   • Recognition  of  an  object  follows  a  ‘match’  b/w  the  stimulus  and  an  internal  template  of  that  object.   • Detect  patterns  by  matching  visual  input  with  a  set  of  templates  stored  in  memory   –  see  if  any  template  matches.   • What  if  objects  differ  slightly  from  template  (i.e.  rotated  or  scaled  differently)?  Use  a  set  of  transformations  to  best  align   the  object  with  a  template  (i.e  rotation,  translation,  scaling…).   • Template  matching  calculates  a  correlation  between  the  input  and  the  template  to  indicate  degrees  of  fit.  However,  the   visual  system  is  capable  of  recognizing  a  huge  variety  of  stimuli  that  would  not  pass  such  test  (i.e.  The  letter  A  in   different  fonts).     • Assumes  an  exact  match  (i.e.  w e  should  have  a  different  template  for  the  letter  A  in  each  font ).   • Requires  a  huge  #  of  templates  to  account  for  all  the  possible  variations  of  objects   • Works  in  highly  constrained  situations   • Well-­‐suited  for  machine  implementation   • Cannot  handle  the  complexity  of  our  everyday  perceptions,  so  can’t  be  the  complete  story     Feature  Theory   • All  objects  are  composed  of  separable,  distinct  parts  –  features  (i.e.  vertical  lines,  horizontal  lines,  circles...)   • Physiological  evidence  for  features :   o Recordings  from  neurons  –  edge  detectors,  movement  detectors,  colour  detectors,  etc.   o Cells  respond  to  very  specific  simple  features   • Cognitive  evidence  for  features:   o Conjunction  search  –  People  are  slower  at  visual  search  when  they  have  to  search  for  a  conjunction  of  features,   instead  of  just  a  single  feature.   o Search  asymmetries  –  Easy  to  find  a  tilted  line  against  vertical  ones,  hard  to  find  a  vertical  line  against  tilted   ones.  It’s  easier  to  find  “tilt”  than  “not  tilt”,  indicating  that  a  feature  we  use  is  “tilt”  not  ”absence  of  tilt”.   • Features  are  primitive  (basic)  to  perception   –  single  features  can  “pop  out”  from  the  background.  When  there’s  a  higher   level  of  analysis  required  (conjunction  search),  there  is  no   pop-­‐out.   • Is  it  enough  to  consider  simple  features?  Can  we  also  acco unt  for  spatial  relations  b/w  components?     Structural  Theory  (i.e.  Recognition  by  Components)   • Includes  both  info  about  components  and  spatial  relations  (how  the  parts  connect).   • The  features  are  defined  as  geometric  components.   • Recognition  by  Components  Theory   o Our  mental  models  of  objects  are  made  up  of   geons  (i.e.  cubes,  cylinder,  wedges,  etc.).  36  posited  geons.   o Objects  are  created  by  combining  geons  by  their  edges.  For   each  object  the  relations  b/w  different  geons  are   specified  (top-­‐of,  side-­‐connected,  etc.)   o Geons  can  be  recognized  from  any  angle,  therefore  an  obj ect  can  be  recognized  from  any  angle  as  long  as  you   can  recognize  the  geons.   o Objects  are  recognized  by  identifying  their  constituent  geons  and  relations  among  them.   § Colour,  shape,  and  texture  are  not  necessary   § All  possible  combos  of  3  geons  and  relations  b /w  them  yields  1.4  billion  possible  3 -­‐geon  objects.   § Biederman  claims  we  use  about  30,000  and  have  names  for  only  3,000   o Evidence  for  geons   § Easier  to  identify  objects  when  geons  are  intact,  than  when  geons  are  disrupted  (even  given  the  same   total  amount  of  info)  (figure  on  right).   § If  you  can’t  recognize  the  geons,  you  can’t  recognize  the  object.   o Our  ability  to  recognize  geons  is  positively  correlated  with  our  ability  to  recognize  objects.   o Priming  involves  an  increased  sensitivity  to  particular  stimuli  as  a  r esult  of  previous  experience.   § Priming  occurs  for  successive  pictures  that  contain  the  same  geons.     § Priming  does  not  occur  for  successive  pictures  that  do  not  contain  the  same  geons.   o Problem  with  geons   § Some  things  are  difficult  to  distinguish  based  on  geons  (i.e.  a  dog  vs.  a  wolf).   § Some  things  would  require  geons  to  be  posited  (i.e.  a  loaf  of  bread)   –  the  point  of  geons  is  to  be   universal,  not  specific.   Word  Recognition   • Limited  set  of  line  segments  (verticals,  horizontals,  curves,  etc.)  make  up  letters.  Lett ers  make  up  words.   • How  to  study  reading?   o Have  to  make  it  harder  (i.e.  make  it  faster,  mask  it).  People  are  usually  too  good  at  it  to  see  any  effects.   • How  to  measure  reading?   o Recognition  threshold  (ms)   o Percentage  recognized     What  makes  word  recognition  easy  or  hard?   • Word  Frequency   o People  are  faster  to  recognize  “happy”  than  they  are  “harpy”.  Not  just  that  people  aren’t  fam
More Less

Related notes for PSYCH 2H03

Log In


OR

Join OneClass

Access over 10 million pages of study
documents for 1.3 million courses.

Sign up

Join to view


OR

By registering, I agree to the Terms and Privacy Policies
Already have an account?
Just a few more details

So we can recommend you notes for your school.

Reset Password

Please enter below the email address you registered with and we will send you a link to reset your password.

Add your courses

Get notes from the top students in your class.


Submit