際際滷

際際滷Share a Scribd company logo

Tervo: Sensory Dissonance Models

Tommi Himberg
Tommi Himberg

There are two aspects of dissonance perception: learned/top-down and innate/bottom-up. Sensory dissonance can be modeled using either auditory models based on the auditory periphery or curve-mapping models based on empirical data. Computer programs that simulate sensory dissonance processing can estimate the degree of dissonance for a given sound. The models were tested on piano music, drone music, and synthesized chords by comparing their predictions of dissonance to participant ratings. The curve-mapping models predicted ratings reasonably well for isolated chords and drone music but not piano music, possibly due to non-sensory influences on ratings for more complex music.

1 of 26
Sensory Dissonance Models Tuukka Tervo - Colloquium 29.10.2009
Sensory Dissonance Models There are two aspects of dissonance perception 1. learned or top-down or contextual 2. innate or bottom-up or sensory
Sensory Dissonance Models Sensory dissonance is explained in terms of Physical properties of sound Physiological properties of the auditory system
Sensory Dissonance Models Computer programs that simulate the sensory process of dissonance perception. give an estimate of the degree of perceived dissonance of a given sound.
Sensory Dissonance Models Two types Auditory models Curve-mapping models
Sensory Dissonance Models Auditory models Based on models of the auditory periphery e.g. Leman (2000)
Sensory Dissonance Models Curve-mapping models Based on empirical data from Plomp and Levelt (1965) e.g. Sethares (1999), Vassilakis (2001)
Curve-mapping Models From Plomp & Levelt 1965 Sensory dissonance of a sine tone pair as a function of frequency difference on a critical bandwidth scale
Sensory Dissonance Models Why try to model sensory dissonance perception? To gain better understanding about its contribution to the organisation of music. May be useful for MIR tasks. May be useful for studies of higher-level processing of music, e.g. music-induced emotions.
Research Question Can the models of sensory dissonance predict the perceived degree of dissonance of music?
Method ! Listening experiment to gather behavioural data on dissonance perception. ! Simulating sensory dissonance processing using various models. ! Statistical analysis of the relation between the models' predictions and the behavioural data.
Participants Two groups of 16 Students of musicology and music education at the University of Jyv辰skyl辰
Stimuli A. Piano music (Keith Jarrett) 50 x 5 seconds B. Drone music (Jim ORourke, Phill Niblock) 50 x 5 seconds C. Synthesized chords 20 x 3 seconds
Procedure Each stimulus is rated on a scale from 1 (consonant) to 7 (dissonant). Group 1 First stimuli A, then stimuli B or vice versa Group 2 Stimuli A and B mixed, then stimuli C
Calculating Sensory Dissonance Models implemented in Matlab Sethares' (1999) and Vassilakis' (2001) curve-mapping models in the MIRtoolbox at the Finnish Centre of Excellence in Interdisciplinary Music Research, University of Jyv辰skyl辰 Leman's (2000) auditory model in the IPEMtoolbox at the Institute for Psychoacoustics and Electronic Music research center of the Department of Musicology at the Ghent University
Calculating Sensory Dissonance 5 seconds of audio
Calculating Sensory Dissonance 50 ms frame
Calculating Sensory Dissonance Spectrum of the 50 ms frame
Calculating Sensory Dissonance Peak-picking
Calculating Sensory Dissonance Roughness of each frame
Calculating Sensory Dissonance Mean roughness
Correlation between predictions and ratings Chords r = 0.7663 p < 0.01
Correlation between predictions and ratings Drone r = 0.5790 p < 0.01
Correlation between predictions and ratings Piano r = 0.2708 p > 0.05
Some Conclusions Curve-mapping models can predict the perceived dissonance reasonably well for ! isolated chords. ! drone music. Dif鍖culties with piano music. Why? Non-sensory aspects affect the ratings? Sharp attacks cause the models to detect erratic dissonance peaks?
References Leman, M. 2000. Visualization and Calculation of the Roughness of Acoustical Musical Signals Using the Synchronization Index Model (SIM). Proceedings of the COST G-6 Conference on Digital Audio Effects. Retrieved from: http://profs.sci.univr.it/~dafx/Final-Papers/ pdf/Leman_DAFXFinalPaper.pdf Plomp, R. & Levelt, W. J. M. 1965. Tonal Consonance and Critical Bandwidth. Journal of the Acoustical Society of America, 38, 548-560. Sethares, W. 1999. Tuning, Timbre, Spectrum, Scale. Berlin, Heidelberg, New York: Springer-Verlag. Vassilakis, P. N. 2001. Perceptual and Physical Properties of Amplitude Fluctuation and their Musical Signi鍖cance. Los Angeles: University of California. Doctoral dissertation.

More Related Content

Tervo: Sensory Dissonance Models

  • 1. Sensory Dissonance Models Tuukka Tervo - Colloquium 29.10.2009
  • 2. Sensory Dissonance Models There are two aspects of dissonance perception 1. learned or top-down or contextual 2. innate or bottom-up or sensory
  • 3. Sensory Dissonance Models Sensory dissonance is explained in terms of Physical properties of sound Physiological properties of the auditory system
  • 4. Sensory Dissonance Models Computer programs that simulate the sensory process of dissonance perception. give an estimate of the degree of perceived dissonance of a given sound.
  • 5. Sensory Dissonance Models Two types Auditory models Curve-mapping models
  • 6. Sensory Dissonance Models Auditory models Based on models of the auditory periphery e.g. Leman (2000)
  • 7. Sensory Dissonance Models Curve-mapping models Based on empirical data from Plomp and Levelt (1965) e.g. Sethares (1999), Vassilakis (2001)
  • 8. Curve-mapping Models From Plomp & Levelt 1965 Sensory dissonance of a sine tone pair as a function of frequency difference on a critical bandwidth scale
  • 9. Sensory Dissonance Models Why try to model sensory dissonance perception? To gain better understanding about its contribution to the organisation of music. May be useful for MIR tasks. May be useful for studies of higher-level processing of music, e.g. music-induced emotions.
  • 10. Research Question Can the models of sensory dissonance predict the perceived degree of dissonance of music?
  • 11. Method ! Listening experiment to gather behavioural data on dissonance perception. ! Simulating sensory dissonance processing using various models. ! Statistical analysis of the relation between the models' predictions and the behavioural data.
  • 12. Participants Two groups of 16 Students of musicology and music education at the University of Jyv辰skyl辰
  • 13. Stimuli A. Piano music (Keith Jarrett) 50 x 5 seconds B. Drone music (Jim ORourke, Phill Niblock) 50 x 5 seconds C. Synthesized chords 20 x 3 seconds
  • 14. Procedure Each stimulus is rated on a scale from 1 (consonant) to 7 (dissonant). Group 1 First stimuli A, then stimuli B or vice versa Group 2 Stimuli A and B mixed, then stimuli C
  • 15. Calculating Sensory Dissonance Models implemented in Matlab Sethares' (1999) and Vassilakis' (2001) curve-mapping models in the MIRtoolbox at the Finnish Centre of Excellence in Interdisciplinary Music Research, University of Jyv辰skyl辰 Leman's (2000) auditory model in the IPEMtoolbox at the Institute for Psychoacoustics and Electronic Music research center of the Department of Musicology at the Ghent University
  • 16. Calculating Sensory Dissonance 5 seconds of audio
  • 17. Calculating Sensory Dissonance 50 ms frame
  • 18. Calculating Sensory Dissonance Spectrum of the 50 ms frame
  • 19. Calculating Sensory Dissonance Peak-picking
  • 20. Calculating Sensory Dissonance Roughness of each frame
  • 21. Calculating Sensory Dissonance Mean roughness
  • 22. Correlation between predictions and ratings Chords r = 0.7663 p < 0.01
  • 23. Correlation between predictions and ratings Drone r = 0.5790 p < 0.01
  • 24. Correlation between predictions and ratings Piano r = 0.2708 p > 0.05
  • 25. Some Conclusions Curve-mapping models can predict the perceived dissonance reasonably well for ! isolated chords. ! drone music. Dif鍖culties with piano music. Why? Non-sensory aspects affect the ratings? Sharp attacks cause the models to detect erratic dissonance peaks?
  • 26. References Leman, M. 2000. Visualization and Calculation of the Roughness of Acoustical Musical Signals Using the Synchronization Index Model (SIM). Proceedings of the COST G-6 Conference on Digital Audio Effects. Retrieved from: http://profs.sci.univr.it/~dafx/Final-Papers/ pdf/Leman_DAFXFinalPaper.pdf Plomp, R. & Levelt, W. J. M. 1965. Tonal Consonance and Critical Bandwidth. Journal of the Acoustical Society of America, 38, 548-560. Sethares, W. 1999. Tuning, Timbre, Spectrum, Scale. Berlin, Heidelberg, New York: Springer-Verlag. Vassilakis, P. N. 2001. Perceptual and Physical Properties of Amplitude Fluctuation and their Musical Signi鍖cance. Los Angeles: University of California. Doctoral dissertation.