Download presentation
Presentation is loading. Please wait.
1
Pinpointing the Beat: Tapping to Expressive Performances. By S. Dixon and W. Goebl (2002). (beat tracking & auto-transcription) Emilios Cambouropoulos (LBDM)
2
Outlines Overview Related work Experiment Setup Result I ~VI Discussion & Conclusion
3
Overview Tapping along with music Beat tracking Many models assume that beats correspond to the onset Proposed Hypothesis: The perception curve is smoother than the onset curve
4
Related Work Finger tapping: constant rate vs synchronize (Madison, 2001) Negative synchronization error: participants tend to tap earlier than the stimulus, about -20 and -60 ms (Wohlchlager & Koch, 2000) What metrical level and at what phase listeners tend to synchronize with the music (Parncutt, 1994; Drake, Penel, Bigand, 2000; Snyder & Krumhansl, 2001) Pianists tapped to different expressive performances. They tend to underestimate long IBIs, compensating for the error on the following tap (Repp, 1999)
5
Experiment Setup Subject: 25 Musically trained participants Average of 19 years for playing their instruments Stimuli – 4 excerpts, Each excerpt repeated 10 times K331:1K284:1K281:3K331:1 ……………… 2-5 secs random duration gaps Total duration 13 minutes 45 seconds
6
Stimuli – one excerpt Label K331:1 Sonata K331 Movt 1st Bars 1-8 Dur. 25s p-IBI 539ms ML 1/8 2nd bar Metrical Level p-IBI = (performed IBI)
7
Experiment Procedure K331:1K284:1K281:3K331:1 ……………… => Recorded as t-IBI (tapped IBI) Allowed participants to practice Beats with no corresponding played notes were interpolated linearly Matched each tap to the nearest played beat time Deleting taps 40% more than the average p-IBI Matched to a beat already had a nearer tap
8
Results - I Metrical levels are expressed as multiples of the default level Two participants tapped on the 2 nd and 4 th quarter note beats of the bar 3 failed tapping relate to participants tapping inconsistently – changing phase during the excerpt
9
Results - II Initial synchronization time: the first 3 successive beats which matched the calculated ML and phase For each excerpt, tapers were able to synchronize on average by the third or fourth beat of the excerpt Average synchronization time
10
Results - III Solid line: t-IBI Dotted line: p-IBI The t-IBIs describe a smoother curve than the p-IBIs t-IBIp-IBI K281:3K331:1
11
Results - IV To test the smoothing hypothesis more rigorously The distance was measured by the RMS difference of the corresponding taps and beats (Cambouropoulos et al., 2001) K331:1, The tempo curve is highly irregular due to relatively long pauses, which are used to emphasis the phrase structure Based on tap-beat time (Time point) Based on t-IBI and p-IBI (Interval)
12
Results - V To find the time lag between tempo changes and changes in tapping rate The lag of 1 tap is most common => participants respond to a tempo change on the tap after it occurs Analysis of time lags of responses to tempo changes, measure of correlation of t-IBIs and p-IBIs
13
Results - VI It was expected that with repetition, the lag would decrease With increasing repetitions, the 0 lag has the best correlation more frequently For K284, no learning trend is seen Authors ’ comment: “ it was like a chamber music rehearsal – you get it right after the third time ” Analysis of time lags of responses to tempo changes – learning effect
14
Discussion & Conclusion Perceived beat is smoother than the played notes Anticipation: on-beats are perceived as anticipating Minimal deviation: the perceptual system minimizes the deviation from strictly metrical time The above implies that timing fluctuations are not necessarily perceived as tempo changes Tapers underestimate timing changes Two related test: Listener preference test Offline beat marking task
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.