Midterm Flashcards
Perception is a result of:
available physical energy
sensitivities of our sense organs
information processing in our brain
What is the “input” and output” of human vision?
Distal stimulus (outside image, 3D) -> proximal stimulus (retinal image, 2D) -> Visual percept, 3D
What are qualities the eye looks for in an image?
Angle, Shape, Size, Lightness and brightness
Fundamental problem of perception:
Every proximal stimulus is consistent with many different distal stimuli.
Optics
The mapping of the 3D scene to the projected image
Inverse optics:
mapping of the projected image to the 3D scene
levels of analysis of perception
- What problem is it solving? (computational analysis)
- What strategy is it adopting? (algorithm)
- How is it implemented in hardware? (brain circuits)
multiple approaches to sensation and perception
- Theoretical (computational)
- Psychological (behavioral)
- Biological (neuroscience)
Psychophysics
Study of relationship between physical world and “psyche” (Gustav Fechner)
Absolute threshold
Minimum intensity needed to evoke a sensation - Boundary between undetectable and detectable
Difference threshold
Minimum change in intensity that leads to a noticeably different stimulus. boundary between “look the same” and “look different”
Weber’s law
Difference threshold is proportional to stimulus intensity ^I = K . I
K = “weber fraction”
Each difference threshold corresponds to a
just noticeable difference (JND)
Method of constant stimuli
fixed set of stimuli
undetectable to easily detectable
Presented multiple time in random order
Respond: YES or NO
Plot “percentage of detections”
Ideal case: 100% detections at and post the absolute threshold
What actually happens: More of a ramp, take 50% as absolute threshold
Plot graph from intensity and proportion of “yes” responses
Method of limits
Fixed set of stimuli Start with weak (undetectable) stimulus Gradually increase intensity Mark "crossover point" Threshold = mean of crossovers
Method of adjustment
Intensities not fixed in advance
Interactively adjusted by observer
Some concerns: No “right answer”, differences in individual criterion/motivation level
Forced-choice methods
Set up task so there’s always a right answer
Example: Dim light flashes either on left or right of screen
If invisible, observer has to guess
If clearly visible -> Accuracy ~100%
75% point is threshold, scale starts at 50
Doctrine of specific nerve energies
What matters is which nerves are stimulated, not how they are stimulated (Johannes Muller)
Lesion studies
Locus of lesion loss in performance
Example: Damage to area MT and motion-blindness
Difficulty in interpretation: correlation does not imply causation
ex. 1: economy of san francisco / golden gate bridge
Single-cell recording
Measure electrical activity from a single neuron, using a microelectrode
Neurons
Cells that integrate and transmit signals
Dendrites
Collect chemical signals
Convert into electrical activity
cell body
integrates electrical activity
Generates nerve impulses
axon
Transmits nerve impulses
terminals
Convert impulse to chemical signals
Pass on other neurons
action potential
“firing” of a nerve impulse
All-or-none
Travels from cell body to terminals
1. Resting potential = -70mV
2. Given sufficient +ve charge (“depolarization”), a sudden upsurge is generated.
3. Spike travels along axon
4. Dies down but overshoots RP before returning
Synapse
Small gap between pre-synaptic and post-synaptic neurons
Neurotransmitters sent across synapse
Modify likelihood of post-synaptic neuron firing
Two kinds of synapses:
Firing of the pre-synaptic neuron either…
increases chances of post-synaptic neuron firing (excitatory synapse) +ve charge (depolarizing)
decreases chances of post-synaptic neuron firing (inhibitory synapse) -ve charge (hyperpolarizing)
The Rate Law
One impulse is not the basic element of information
Continuous information is encoded by rate of firing
(Hertz = # per second)
What counts as “no response”?
Baseline firing rate: ~1-5 hz
Excitation increases firing rate (100-500 hz)
Inhibition decreases firing rate (< 1 hz)
Firing rate is always measured relative to baseline
EEG
Record brain activity using electrodes on scalp
Difficulties:
(a) Hard to pinpoint precisely
(b) Many signals too weak
Neuroimaging
Highly active regions will have greater blood flow - PET, fMRI
To understand visual perception, we must study
- Light and its interaction with objects
- Structure and function of the eye
- Information processing in the eye and brain
Light
Dual nature: Light is a particle and a wave
Light as particles
Travels in straight lines “rays”
Smallest ‘packet’: Photon
Light as wave
Has a wavelength
Refraction: bends when it encounters a new medium
What is an eye?
Def 2: a structure/organ that can compare light from different directions
convex lens
Convex lenses converge light rays
Focal length
Distance at which parallel rays converge
Diopters = 1 / focal length (in m)
Example: 5 diopters: f = 20cm (=1/5 m)
Main functions of the human eye
Main functions:
Form a sharp image
Transduction
Initiate image processing
Optical power is made up of
cornea (2/3) + lens (1/3)
Optical power of the lens in adjustable (Ciliary muscles, known as accomodation)
Hyperopia (Farsightedness)
Eyeball too short or lens too weak
nearby objects are blurred (rays do not converge enough)
Correction: Convex lens
Myopia (Nearsightedness)
Eyeball too long or lens too “strong”
Distant objects are blurred (rays converge too much)
Correction: concave lens
Presbyopia (“old sight”)
Lens becomes inflexible
Cannot focus on nearby objects
Near point: closest distance at which an object can be focused.
The retina
Light -> Ganglion ->Bipolar -> horizontal -> rod in physical direction
Ganglion cell axons bond to form the optic nerve
Fovea
Small central “pit where vision is most acute”
Optic disk
Where the ganglion fibers leave the eye
rods
Higher sensitivity to light
Lower resolution
Scotopic vision
Color-blind
cone
Lower sensitivity to light
Higher resolution
Photopic (color) vision
Distribution of receptors
Fovea: Cones only -> high-resolution vision
Periphery: rods and cones
How does light at different locations affect a ganglion cell?
- Most of the retina - “no response”
- Small circular region where light excites the cell - ON response
- Donut-shaped region where light inhibits the cell - OFF response
Receptive field
that region on the retina which, when stimulated, influences the baseline firing rate of a neuron. Combination of disk and ring = receptive field
Center-surround antagonism or lateral inhibition
when neuronal activity antagonizes (turns off) surrounding activity (center-surround antagonism)
What do ganglion cells respond to?
- Uniform illumination: + and - responses cancel
do not respond well to overall light level. - Dark-light border: strong response
Respond to changes in light level - Orientation change: no influence
not sensitive to edge orientation
P-cells
parvocellular (small)
Comprise 80% of cells
Smaller receptive fields -> higher spatial resolution
Lower sensitivity
Thinner axons -> worse temporal resolution
color sensitive
M-cells
Comprise 10% of cells
Larger receptive fields -> lower spatial resolution
higher sensitivity
thicker axons -> better temporal resolution
Color blind
Why respond to changes in brightness?
That’s where there is the most information in a scene.
Perceptual consequences of ganglion-cell processing
- Neural signal depends on local intensity and surrounding intensity
- Signal emphasizes contrast borders; de-emphasizes homogeneous regions
Optic chiasm
Temporal half of retina -> Ipsilateral visual cortex
Nasal half of retina -> Contralateral visual cortex
Why does the optic chiasm split the visual field as it does?
Because the controlateral brain areas correspond to the eyes’ visual field - the hemispherical set-up helps establish 3D vision both binocularly, and monocularly.
Left visual fields (both eyes) -> right visual cortex
Right visual fields (both eyes) -> Left visual cortex
Retinotopic map in V1
Each hemisphere represents contralateral visual field
Cortical magnification
80% of cells devoted to central 10 degrees
Main new features of V1 cells
orientation selectivity
Selectivity for direction of motion
Binocularity
Simple V1 cells
respond to edges and bars of specific orientations
Elongated RFs with clearly-demarcated ON and OFF regions
The edge or bar much be positioned exactly within RF
Complex V1 cells
also orientation selective
No separate ON / OFF regions
Exact positioning of edge / bar not required
Binocularity
First site of binocular cells
Note: these cells have two receptive fields!
Ocular dominance
slightly stronger responses to one eye
Example: pattern, looks like a fingerprint
Black stripes: right-eye dominant
White stripes: left-eye dominant
Where does the signal go from V1?
Dorsal stream goes towards parietal (M pathway)
Ventral stream goes towards temporal (P pathway)
Spatial vision
ability to visually detect spatial patterns
example: seashells at multiple scales, zooming out looks like mona lisa
How good is our vision at different scales?
Relation of RF size?
Large RF’s -> coarse scale
Small RF’s -> fine scale
Multi-channel model
Campbell & Robson (1968)
The visual system analyzes information through multiple channels
Each channel is responsible for a particular spatial scale
Fourier’s theorem
A mathematical procedure by which any signal can be separated into component sine waves at different frequencies. Combining these sine waves will reproduce the original signal.
Fechner’s law
A principle describing the relationship between stimulus and resulting sensation that says the magnitude of subjective sensation increases proportionally to the logarithm of the stimulus intensity.
spatial frequency
The number of cycles of a grating (e.g., dark and bright bars) per unit of visual angle (usually specified in cycles per degree).
Why use sine gratings?
1. Spatial frequency How many cycles per unit distance? 2. Amplitude / contrast low: dim, gets darker as you go from low to high 3. Orientation 0, +45, -45, 90 4. Phase Position relative to a fixed landmark
Contrast-sensitivity function (CSF)
A function describing how the sensitivity to contrast (defined as the reciprocal of the contrast threshold) depends on the spatial frequency (size) of the stimulus.
Contrast threshold
Minimum amount of contrast (on a sine grating) that is visible
How low can you go?
Measuring the CSF
Pick a frequency f. Measure contrast threshold for f. Sensitivity = 1/threshold repeat for different frequencies CSF charts = x = spatial frequency, y=contrast sensitivity CSF= window of visibility
Selective adaptation experiment
- Measure an observer’s CSF
- Adapt the observer to a high-contrast grating with some frequency f
- Measure the CSF again
- compare pre- and post-adaptation CSFs.
Adaptation
decrease in the strength of a neuron’s response after prolonged firing
Selective adaptation
Only those neurons sensitive to the adapting frequency get fatigued
Different channels respond to different frequency ranges
Compare vision across different conditions:
Scotopic (low), Mesopic (medium), Photopic (high)
Perception
Our link and access to our world
Construction of our reality
sense of 3D space/distance
sounds/voices
tactile sensations
Perception informs an organism about: what is in its environment and where it is
Evolutionary significant actions: flee from predators, hunt/gather food, find mates, navigate