Pattern Recognition Nick Lund Attention and Pattern Recognition
Introduction Pattern recognition has been defined as ‘the ability to abstract and integrate certain elements of a stimulus into an organised scheme for memory storage and retrieval’ (Solso,1998).
Features of pattern recognition Five principles(Solso,1998): 1.Quickly and accurately. 2.Recognise and classify unfamiliar objects. 3.Accurately recognise shapes and objects from different angles.
Features of pattern recognition 4.Identify patterns and objects even when partly hidden. 5.Recognise patterns quickly,with ease,and with automaticity.
Features of pattern recognition Template matching theories Feature analysis Prototype theories
Top-down and bottom-up processing An examination of the theories of pattern recognition raises the question of whether pattern recognition involves top-down or bottom-up processing.
Bottom- up process Template matching theory Geons (Structural- description theory) Feature theory Prototype theory
Template matching theory External stumuli matches internal template Vast numbers of templates are stored Templates are created by experience E.x.: visit Russia
A A A A Problems: 1) Only occurs when there’s a one-to-one match Payne and Wenger (1998)
2) Where are all these templates stored? 3) Slow process Ex.: recognize 1000-1500 letters a minute 4) To recognize new variations of a pattern (Solso, 1998)
Geons (Structural- description theory) Biederman (1987) Limited number (24) of simple geometric shapes , or geons for us to analyse patterns. Ex.: mug v.s. bucket
Advantage: Recognize pattern from different angles Disadvantage: Can’t explain why we recognize a particular chair. Ex.: My face v.s. my friend’s face
Feature theory Patterns are recognized by analysis of the individual features of the pattern. Ex.: MANGO M A
Four stages of pattern recognition Image demons: record the image and pass it Feature demons: analyse the image for specific feature Ex.: Pandemonium model (Selfridge, 1959)
Cognitive demons: detect the present feature and shout Decision demons: pick up the loudest feature
Advantage: More flexible than template theory Ex.: A , regardless of size, shape or orientation Neisser: M -> N H M V -> longer RT (more distracter) M -> O M Q G -> shorter RT
Disadvantage: Fail to account for the effects of context and expectations Ex.: Two experiments which can’t explain feature theory Fail to detect “t” in “the” (Healy) Teacup, eyebrow -> cup, eye ; eac, ebr (Inhoff and Topolski)
Eysenck and Keane:
How can we recognize patterns when their features are hidden from view?
Relationship between the features
Ex. : T v.s. L
The biology of feature theory Hubel and Wiesel (1959) In visual cortex: Simple cells: particular orientation, specific location complex cells: lines or edges, in all visual field Hypercomplex cells: length and angles, combination of features
Disadvantage: Some questioned the existence of hypercomlex cells How specialized they need to be? Run out of cells
Prototype theory External stimuli matches with internal abstract prototype Ex.: R, compare to all other “R”s Solso(1998): two theoretical models Central-tendency model:
The average, mean
Most common combination, mode
Advantage: More “economical” than template theory Don’t need templates for every shape, size, orientation Explain the speed of recognition of letters, words and patterns / novel stimuli
Disadvantage: can’t explain the effect of context (Eysenck, 1993) old woman v.s. young lady B, R, P have similar prototype
Top-down processing Word superiority effect Rat-Man demonstration