Perception - Week 2

The importance of perception

• Perception and sensation are two different concepts.

• Sensation strictly refers to physical energy conversion to neurosignals.

• Perception is the interpretation of sensory information/neurosignals.

Processes behind perception

• Sensory inputs are converted into perceptions of desks and computers, flowers and buildings, cars and planes, into sights, sounds, smells, tastes, and touch experiences.

• A particuldar problem for psychologists is explaining how sensory organs process the physical energy they receive, laying the foundation for perceptual experiences.

• Multi-sensory perception matters, however, to understand perception best is to look at one sense at a time.

• Visual perception feels ‘easy’, but it’s very complex.

• Only until recently scientists developed systems to match human perceptual abilities.

• Machine systems stuggle with basic object recognition.

• Visuals scenes are complex and have lots of visual information, often overlapping, making it difficult to segregate the information.

• Objects also need to be recognisable from different viewpoints.

• Computers have exceeded human visual abilities to some aspects, computer visions is better in terms of image recognition, surrevilance, detecting and tracking in real-time. For instance it is used insurrveilace of criminals, however, many people of colour are targeted based on previous records, thus raising ethical concerns are prejudices are reinforced

• Recognise objects from different viewpoints.

• Recognise objects from ambiguous information.

• Visual perception relies on the eyes (our sensory organs)

  For example:

  
  

• Objects coming towards us as a bigger retinal image, we understand the objects remain the same size (size constancy) just distance is changed.

• Size-distance scaling hypothesis: size illusions

  • Biased to assume size constancy

  • If the usual size of an object is known, size of retinal image gives distance.

  • However, assumptions can be wrong!

    

• Assume (reasonably, based on experience) that the room is cuboid with walls at right angles to one another.

• Retinal images of people are different sizes

• Ability to determine distance is removed

• When objects are further away, objects appear smaller in retinal image

  • Person A is much further away than B

  • Does not conform to apparent shape of room

  • Make assumption person A is much smaller than person B

• Visual illusions occur when our visual systems makes an error, often because of heuristics (assumptions and expectations) during signal interpretation.

• Similarly, by identifying and analysing the errors made by the visual system in an illusion, we get to understand how the vidual system makes inferences in a normal context.

• Models of visual perception are not based only on illusions (though illusions are widely used in experiments!)

  

Models of visual perception (object recognition)

• All theories share common ground in their exploration of perceptual processes and the way humans make sense of sensory information.

• All theories share common ground in their exploration of perceptual processes and the way humans make sense of sensory information.

• All theories acknowledge that perception is not a passive process of recording stimuli but involves complex cognitive processes that shape our perception of the world.

• Theories differ in how many perceptual processes they can explain.

  • Likelihood Principle (Helmholz)

  • Srreucturalist Approach

  • Gestalt Laws

  • Bayesian Inference

Theory of unconcious inference (isolated perceptual component)

• Likelihood principle (Helmholz)

  • One of the first theories on how we perceive.

  • Helmholz ignored by many of his contemporariws, but greatly influencing modern theories of perception.

  • Focused on non-concious perception.

  • Perceive the object that is most likely given the situation (using knowledge and experience)

Structuralist Approach (isolated perceptual components)

• Perceptions are created by combining elements called sensations.

• Perception of form produced by combining these elemental components.

• Elemental components were established by asking participants to use ‘analytical introspection’ but there was little consenseus.

• Can’t explain illusory contours or ambiguous figures

• By combining different elements we can create our own world, meanings and images. Wundt called them sensations.

• Wundt proposed we perceive the environment by combining all out elements together. However this can’t explain illusory contours hence structural approach cant satisfy what happens in our visual system.

Gerstalt Laws (whole more than the sum of its parts)

• The Gestalt Approach (heuristics - “best guess rules”)

• Emphasize the idea that the mind organizes stimuli into wholes , rather than perceiving isolated elemtents.

• .Argue that perception is not simply the sum of individual sensory stimuli (like in the theory of unconscious inference of structuralism), but involves inherent principles of organization, such as proximity, similarity, and closure.

• Fundamental principle is the Law of Pragnanz - typically perceive the simplest possible organisation of visual world.

• Most other principles can be organised under this law

  • Similarity: appear grouped together

  • Good continuation

  • Proximity

  • Closure

  • Common Fate

  • Meaningfulness of Familiarity

• Other Gerstalt Principles

  • Fugure-Ground Segregation

  • Surface or Area

  • Common region

  • Connectedness

  • Synchrony

  • Pareidolia

  • Orientation

  • Symmetry

  • Convexity

Bayesian Inference (probability theory to explain how we perceive)

• Statistical technique restates Likelihood Principle in terms of probabilities (allowing AI or other systems to be ‘programmed’)

• In perceptual terms, if asking participants whether object in a kitched is more likley to be a loaf of bread or a mailbox, the load is more likely and perceptual system concludes that it’s a loaf.

• Probablility of outcome is determined by 2 factors

  • Prior (probability) - initial belief about probability of outcome

  • Likelihood - extent to which evidence is consistent with outcome

• One strenght is its adaptibility. As new sensory information becomes available, the system can continuously update its beliefs about the identity of objects. This adaptibility is crucial for real-world scenarios where conditions may change dynamically.

• Applying Bayesian inference to object recognition helps address the challenges of uncertainty and variablitiy in sensory input, making it a valuable framework for understanding how our brains and artificial systems make sense of the visual world.

Perception for Action

• What/where model is based on our brain structure - ventral/retro is another name.

• Developed by Goodale and Milner

  

…

Textbook notes: