Title: FACE RECOGNITION
1FACE RECOGNITION
- CORE READING
- Ellis, A and Young, A. (1996). Human cognitive
neuropsychology. Psychology Press, Chap 4. - SUPPLEMENTARY READING
- Burton, M., Bruce, V., Johnston, R. (1990).
Understanding face recognition with an
interactive activation model. British Journal of
Psychology, 81, 361-380.
- AIMS OBJECTIVES
- The aim of this lecture is to review models of
face recognition. - At the end of the lecture you will have learned
- Modular view of face recognition
- Mechanisms in face recognition
- Models of face recognition
- What is prosopagnosia
2Adaptive reasons to recognise faces
- Evolutionary reasons
- We need to distinguish between familiar faces in
our social groups for survival. - Socio-cultural reasons
- We might be able to predict a persons behaviour
from facial features.
- Genetic predisposition?
- Brain regions dedicated to recognition of faces
(in primates too). - Damage to this part of the brain leads to a
recognition disorder called prosopagnosia.
3How do we recognise faces?
- When we look at a face we can extract different
types of feature information. - For all faces we can decide upon emotional
expression, gender, age, etc. - This information enables us to distinguish
between faces and identify people.
- Feature information enables us to make
familiarity judgements on the basis of the
physical features of faces. - But this does not ensure that we identify people
correctly - I know your face but Ive forgotten your name
4Face knowledge
- There are several levels of modular knowledge
used in the identification of familiar faces. - Traditional models of cognition assume these are
accessed in stages of information processing.
- 1) Recognising a face as familiar.
- 2) Retrieving our stored biographical knowledge
about a persons identity. - 3) Retrieving name of the person.
5Bruce and Young model (1986)
- Non-identity related information
- Facial expression.
- Directed judgements of age, gender etc..
- Lip reading.
- Identity related information
- Face recognition units (FRUs).
- Person identity nodes (PINs).
- Name generation.
6How do we recognise faces?
- Pattern recognition.
- Structural encoding.
- Perceive and create a visual 'description' of the
features of a face. - shape, size, colour etc and their configuration.
7Stages in recognizing faces
- Face recognition units (FRUs).
- Matching the visual description of a seen face
against stored descriptions of the appearance of
familiar faces seen before.
- Person identity nodes (PINs).
- Access to biographical information held in
long-term memory concerning the familiar person's
occupation, personality, etc.
8Unusual views
Fear, anger
Perception of features of face (shape, size)
Stored representations of familiar faces
Knowledge about familiar people (occupation etc)
Specific names of people
Bruce and Young, 1986
9Evidence for the model
- Experimental evidence
- Familiarity decisions are faster than PIN
decisions (Young et al., 1986b). - Have you seen this face before versus is this an
actor? YES or NO (button press). - Therefore, access to Face Recognition Units must
precede access to Person Identity Nodes
(knowledge).
- Reaction times (RTs) to PIN decisions about faces
are faster than face naming (Young et al. 1986). - Is this an actor versus Is this John
Malkovitch? - Person identity knowledge is accessed before the
stage of name information.
10Prosopagnosia
- An acquired inability to recognize familiar
faces - famous faces
- friends
- family
- own face in the mirror.
- Patients know when they are looking at a face,
but they cannot say who the face belongs to. - They rely on other non-facial cues for person
identification e.g., gait, clothing.
11More evidence for the model
- Neuropsychological
- Prosopagnosia patients
- Dissociations between the recognition of familiar
people (prosopagnosia) and lip reading in hearing
impaired or deaf signers (Campbell et al. 1986).
- Mrs D cannot recognise faces but she can lip
read. - Mrs T can recognise faces but she cannot lip
read. - Impaired naming of faces can occur with intact
access to biographical knowledge about a person. - This is called anomia.
12Evidence against the model
- Implicit recognition of faces by prosopagnosic
patients. - PH could classify famous faces into their correct
occupation even though he did not recognise the
face. - So PIN information can be accessed without
conscious recognition of FRUs.
- De Haan et al (1991) reported a prosopagnosic who
could match faces to the names of famous people
correctly. - However, she was unable to recall any
biographical (PIN) information about the face. - Is this an actor? DK
- So name generation can occur without first
accessing the FRUs of a famous face.
13IAC model of face recognition
- Units within a pool are linked via inhibitory
connections. - There are excitatory connections between
corresponding units in different pools. - Links are bidirectional.
- Model is a sequential cascaded account.
- Information processing at one stage need not be
complete before processing at the next level is
initiated (c.f. Bruce et al., 1990).
14Impaired for PH
Intact for PH
Burton, Bruce Johnston, 1990
15Evaluation of IAC model
- Cognitive models of face recognition must
explain - Familiarity effects.
- Implicit processing by prosopagnosic patients.
- IAC model explains PHs access to PIN knowledge
even though he cannot access FRU knowledge.
- Are representations of faces and their identities
stored as a local units or as distributed units? - The IAC model and the behavioural evidence
suggests representations of faces are stored as
distributed units.
16The fusiform gyrus and vision
Tools
Objects
Faces
17(No Transcript)
18Summary
- Face recognition involves
- Physical information (shape, colour, form).
- Semantic information (biographical).
- Phonological information (names).
- Cognitive models of face recognition can explain
familiarity effects on normal face recognition
and implicit processing of faces in prosopagnosia.