Sei sulla pagina 1di 13

Phase Coherence as a

Measure of Perceptual
Synchrony in Coupled-Oscillator
Networks

Nolan Lem
Stanford University - CCRMA
Motivation
• Perception of temporal regularity is a crucial aspect in our
ability to listen, process, and construct meaning from musical
events and the acoustic ecologies we inhabit.
• Models of network-based synchronization have been
examined in the context of music cognition and biological
dynamical systems.
• Synchrony as an processual, auditory percept
 Time-based percept that arises from a plethora of
competing auditory stimuli.
Summary of background
1. Large and Kolen (1994) – “Resonance and the Perception of
Musical Matter”
 Attempt to model temporal musical expectancy…
 Applied mathematical model of entrainment via phase and frequency locking
coupled oscillators to the incoming rhythmic patterns played by a performer on a
midi piano.
2. Large and Jones (1999) – Dynamics of Attending: How People Track
Time-Varying Events
Dynamic Attending Theory (DAT):
external auditory event’s rhythm  attending rhythms/internal oscillations  temporal expectancy

incorporates attentional focus into model


• attention is dynamic and adapts to changes in attending rhythm and event
contexts
Pattern Context Effects – stimuli within a trial that change over time
affect attentional focus (such as entrainment)
Large, E. W., & Jones, M. R. 1999. The Dynamics of A
Session Context Effects = across trials, Track Time-Varying Events. Psychological Review, 10
Questions and Background
Methods and Materials

• Ten Stanford University Students


 Did not take into account musical training
• Stimuli
• Nine audio tracks exhibiting different
pattern contexts (generated via ChucK)
1. Ramp (three trials @ 2 min)
 Phase coherence ramped up and down
 @ mean synchrony frequencies:
 ⍵l = 0.67 Hz, ⍵m = 2.28 Hz, ⍵h =9.5 Hz

2. Interval (three trials @ 1 min)


 Phase coherence jumps around
3. Density (three trials @ 1 min)
 20, 40,150 oscillator populations
Methods and materials con’t…
Task:
Listener asked to move digital slider in response to each audio
track in concert with what they consider to be ’not synchronous at all’
 ‘moderately synchronous’  ‘extremely synchronous’
Expected results based on the
hypotheses
• Average Synchrony Contour from participants would
follow the phase coherence contour as a function of
time
• Pattern context effects would be notable
 Pulse percept would entrain listeners over the course of
individual trials and across trials (session context effects)
Experiment 1:
Ramp Context
Trial 1 -3

Figures 2-[1-3] (left column): Ramp Context Synchrony Contours


vs. Phase Coherence r(n). Figures 2-[4-6]: Perceptual Error Over
time
Experiment 1 con’t:
Intervallic Context
(trial 4-6)
10 second intervals at
⍵m
shaded color regions

Phase coherence jumps at


intervals:
[0.35, 0.18, 0.6, 0.9, 0.21,
0.7]
At
Participants were able to
follow general contour of
phase coherence.
Most accurate ratings at
nearly full synchrony and
non-synchrony Figures 3-[1-3] (left column): Intervallic Context Synchrony
Contours vs. r(n). Figures 3-[4-6]: Perceptual Error Over
time
Experiment 2:
Density Context
Trials 7-9

Number of oscillators in system


changed from 2040150.
Low density, medium density and
high density
Participants as a whole were
more accurate in their synchrony
rating for the high density system.
Less dense system creates
impression of dense polyrhythms
that make it more difficult to
discern higher level features.

Figures 4-[1,2,3] (left column): Density Context Synchrony


Contours vs. r(n). Figures 4-[4,5,6]: Perceptual Error Over time
Total Percent Error relative to entire trial context

Numerical Integration:
Using normalized uniform-grid Ramp Context
trapezoidal function to integrate ⍵l ⍵m ⍵h
error function*, e(n) over each trial 0.208 0.118 0.160
Percent Error
context window. Interval
Context
Gives an indication of ⍵l ⍵m ⍵h
percent error: similarity Percent Error 0.0937 0.0767 0.182
measure of how well Density
Context (⍵m)
average contour aligns Low medium high
with phase coherence Percent Error 0.149 0.132 0.101
contour. Table 1: Total Percent Error – per Trial Context
Discussion
In general, participants’ rated synchrony followed the contour of
the phase coherence over time.

Were more likely to over estimate the synchrony relative to the


phase coherence.
+ positive percent error across trials
RAMP Context Interval Context Density Context
⍵l ⍵m ⍵h ⍵l ⍵m ⍵h ⍵m ⍵m ⍵m
negative 0.0411 0.16118 0.0385 0.237 0.4301 0.06 0.319 0.1737 0.248
positive 0.9588 0.8388 0.9614 0.763 0.5698 0.9399 0.681 0.8263 0.752

Table 2: Rated Synchrony Percent Error Positive vs. Negative

Discuss any problems and potential explanations for


the pattern of the obtained data
Conclusions
what's tested and learned from this experiment

what future research can address

Potrebbero piacerti anche