AHOP - TASK 4
AUDITORY SCENE ANALYSIS

Auditory scene
analysis

No information
which vibration is created by
which source (e.g. instrument)

Auditory scene

array of sound sources
in environment

Auditory scene analysis

Multiple sounds at same time (e.g. symphonic orchestra)

separation of multiple stimuli
produces by each of the sources
in the scene into separate perceptions

deciding which frequency components
belong together to form each sound

Principles of auditory
grouping
(gestalt, cues)

PLACE
same source = same place,
movement (mostly)

TIMBRE, PITCH
same source = same pitch/timbre

ONSET TIME
same source = start same time

PROXIMITY IN TIME
repetition rate rapid = same source

Auditory stream
grouping
(integration)

same source, same onset, same pitch/timbre, close in time (synchrony)

one stream, object

Auditory stream
segregation

two streams, two objects

diff source, diff onset, diff pitch/timbre, far in time (asynchrony)

separation of two
experiments

Tonotopic map
Large diff frequency = non-overlapping

Neural adaption

Forward suppression

ELHILALI

FISHMAN

Behavioural
experiment

Neural correlate

synchrony/coherence is primary
(overrules high freq separations)

physiology

Psychophysics

Computational modeling

thresholds for detection

humans

informal listening

single stream = lower threshold

two streams = higher threshold

last tone slightly shifted temporally

segregation (asynchrony) = hard to detect

Experiment 1

single unit recording (A1)

ferrets

fixed freq separation
(1, 0.5, 0.25 octaves)

Segregation
1, 0.5 octaves

Grouped
(0.25 octaves)

Synchronous = always grouped
(irrespective of delta freq)

synchronous, overlapping
and alternating presentation

No neuronal correlate found
(firing rate same for all presentation modes)

strongest response
BF site

Experiment 2

frequency range
of interactions

A (BF) tone fixed,
B changed

No neuronal correlate found
(firing rate pattern same)

Temporal coherence is
fundamental orgnaising principle

temporal integration

coherence analysis

spatio-temporal hypothesis

PFC? Attention?

Failed to find neuronal correlate

Auditory scene analysis

Auditory systems groups and segregates components of mixtures to construct perceptual representations of sound sources (auditory images)

Gestalt principles

Auditory stream segregation

sequential organisation

one stream

slow PR

small freq separation

Two streams

fast PR

high freq separation

MUA & CSD
complementary techniques

Neural populations

Monkeys

Stimuli

Alternating frequency (AF)

Same frequency (SF)

Best frequency (BF), tone A

non-BF, tone B

A1

Results

Same frequency

Alternating frequency

Reversed order
(no difference, A dominant)

A higher response (BF)

higher PR = decrease response

higher PR = decrease response

B (non-BF) faster decrease

40Hz = only response to A
at half of PR (20 Hz)

Model of neural
stream segregation (A1)

CSD-methods

Source = negative

Sink is positive

Same results as MUA

Attention biases
stream segregation

Intermediate freq
separation

dependent of PR

small freq
separation

overlap

grouped, response both

decrease response fast PR

high freq
separation

no overlap

segregation

decrease amplitude higher PR

preferential processing at 40 Hz

Proposed mechanism

Forward masking
/suppression

suppression of subsequent responses

Source (inhibitory) after
sink (excitatory)

BF tones more effective
in subsequent suppression of non-BF

BF response larger under SF
than AF (less suppression)

BF suppresses next BF more

increase suppression
non-BF with increase separation freq