Loading presentation...

Present Remotely

Send the link below via email or IM

Copy

Present to your audience

Start remote presentation

  • Invited audience members will follow you as you navigate and present
  • People invited to a presentation do not need a Prezi account
  • This link expires 10 minutes after you close the presentation
  • A maximum of 30 users can follow your presentation
  • Learn more about this feature in our knowledge base article

Do you really want to delete this prezi?

Neither you, nor the coeditors you shared it with will be able to recover it again.

DeleteCancel

Play it Again: Evolved Audio Effects and Synthesizer Program

No description
by

Ben Smith

on 19 April 2017

Comments (0)

Please log in to add your comment.

Report abuse

Transcript of Play it Again: Evolved Audio Effects and Synthesizer Program

Play it Again: Evolved Audio Effects and Synthesizer Programming
Benjamin D. Smith
Indianapolis University-Purdue University-Indianapolis (IUPUI)
bds6@iupui.edu http://ben.musicsmiths.us

Reverse Engineering Audio Device and Synth Presets
EC Model
User provided target sound sample
Evolve parameter set for synthesis or filter engine
Goals for Real World Use
Generic operation on any/all devices
Speed:
fast!
Practicalities
Most commercial devices only expose subset of parameters (API)
All Native Instruments devices and synths, Sylenth, etc.
Most commercial devices have 500+ parameters
Choices
Ableton Live
Exposes all parameters, API access
Max for Live for spectral analysis
Previous Work
Modular OP-1, multi-objective GA, 2016
GAs for FM "tone matching" in early 1990s
GA Model
Audio Analysis
MFCC, 1st 13 cepstra, normalized
Adaptive Modifications
Fitness: squared error, MFCC data
Crossover: one-, two-point, uniform performed well
Mutation: Gaussian addition
Selection: roulette wheel, no in-breeding, 4-6% elite
maintain diversity, improve movement through exploration and exploitation behavior
M & C probabilities based on ratio of individual fitness to population maximum and mean
Also use adaptive Selection Pressure coefficient:
~10% crossover
~30% crossover
No privileged knowledge, no optimizations for special parameters, parameter relationships
Easy for typical musician to use in daily creative practice
Why/When does it fail?
MFCC analysis idiosyncrasies?
Problem:
susceptible to small differences
very specific, not general
confused by random/LFO effects
Better efficiency for large parameter devices?
Sub-population clustering for selection?
1. Garcia, R.: Growing sound synthesizers using evolutionary methods. In: Proceedings ALMMA 2001: Artificial Life Models for Musical Applications Work- shop,(ECAL 2001) (2001)
2. Horner, A.: Double-modulator FM matching of instrument tones. Computer Music Journal 20(2), 57–71 (1996)
3. Horner,A.:Nested modulator and feedback fm matching of instrument tones. IEEE Transactions on Speech and Audio Processing 6(4), 398–409 (1998)
4. Horner, A., Beauchamp, J., Haken, L.: Machine tongues xvi: Genetic algorithms and their application to fm matching synthesis. Computer Music Journal 17(4), 17–29 (1993)
5. Johnson, A., Phillips, I.: Sound Resynthesis with a Genetic Algorithm. Imperial College London (2011)
7. Lai, Y., andDer Tzung Liu, S.K.J., Liu, Y.C.: Automated optimization of parameters for FM sound synthesis with genetic algorithms. In: Proceedings of the International Workshop on Computer Music and Audio Technology. Citeseer (2006)
8. Macret, M., Pasquier, P.: Automatic design of sound synthesizers as pure data patches using coevolutionary mixed-typed cartesian genetic programming. In: Proceedings of the 2014 Annual Conference on Genetic and Evolutionary Computa- tion. pp. 309–16. ACM (2014)
9. Macret, M.M.J.: Automatic Tuning of the Op-1 Synthesizer Using a Multi-Objective Genetic Algorithm. Doctoral dissertation, Simon Fraiser University, Van- couver, CN (2013)
13. Tan, B., Lim, S.: Automated parameter optimization of double frequency modulation synthesis using the genetic annealing algorithm. Journal of the Audio Engineering Society 44(1/2), 3–15 (1996)
14. Tatar, K., Macret, M., Pasquier, P.: Automatic synthesizer preset generation with presetgen. Journal of New Music Research 45(2), 124–44 (2016)
16. Yee-King, M., Roth, M.: Synthbot: An unsupervised software synthesizer programmer. In: Proceedings of the International Computer Music Conference. Ireland (2008)
Lack of Audio Effect examples.
Digital Filter
Alg. Generator
Drum Rhythm Sample
Target (with Phaser)
Evolved Solution
Results
Deterministic effects & Synths
(distortion/EQ/compression, "Electric")
89% convergence in < 100 epochs
Stochastic effects
final fitness match:
57.67% mean,
31.85% std
> 99% solution phenotype match
Parameter Tuning Problem
Synths can exhibit low parameter accuracy
(alt. solutions found)
Target Spectra
Solution Spectra
Target
Solution
no EC/GA knowledge
Phenotype
Genotype
*Tested with randomly
generated targets
Randomization of targets not reflective of human usage?
Full transcript