Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Ragabot-Music Encoded

Ragabot-Music Encoded

In this talk we will attempt to encode some of the common structures in Hindustani Classical music and make a ragabot: a code/bot that creates music. This is a fun exploration with some interesting experiments without a claim that we can really play god! May be just a little bit :)

vikrant

July 20, 2019
Tweet

More Decks by vikrant

Other Decks in Programming

Transcript

  1. Harsh Vinjamoor Design of control algorithms for mechatronic systems Masters

    in control systems and PhD in mathematics Working in a start-up that makes control units for engines and motors Linkedin: https://in.linkedin.com/in/harsh-vinjamoor-82482a12
  2. Vikrant Patil @vikipedia Software Consultant, Trainer. A curious Amature in

    musical instruments. Advanced Programming Courses At Pipal Academy. Worked at DRDO and Strand Life Sciences.
  3. Structures in a raga: a brief Sa Re Ga Ma

    Pa Dha Ni SA SA Dha Pa Ga Re Sa Re Ga$ Ga Re dha$ Sa Missing notes and important notes Trajectories ‘Typical’ sequences
  4. Structures in a raga: a simple example Bhoop: relatively straight

    forward ▪ Aaroha: Sa, Re, Ga, Pa, Dha, SA ▪ Avaroha: SA, Dha, Pa, Ga, Re, Sa ▪ Pakad: Ga, Re, dha, Sa--; Ga, Re, Pa, Ga; dha, dha, Sa Sa Re Ga Ma Pa Dha Ni
  5. Structures in a raga: a complicated example Tilak kamod: a

    very vakra raga! ▪ Aaroha: pa, ni, Sa, Re, Ga, Sa---Re, Ma, Pa, Ni,SA ▪ Avaroha: SA, Pa, Dha, Ma, Ga--Sa, Re, Ga, Sa, ni ▪ Pakad: pa, ni, Sa, Re, Ga, Sa---Re, Pa, Ma, Ga, Sa Sa Re Ga Ma Pa Dha Ni
  6. Structures in a raga: encode with Markov chains Sa Ga

    Pa Re Dha Sa Ga Pa Re Dha Sa Ga Pa Re Dha Sa Ga Pa Re Dha Sa dha dha dha dha Ga Pa Re Dha Sa dha • Missing notes have zero probability • Important notes have higher probability • Trajectories governed by probabilities • Characteristic phrases are more probable Sa Re Ga Pa Dha SA Sa 0.2 0.2 0.1 Re 0.4 0.1 0.3 Ga 0.3 0.4 0.3 Pa 0.1 0.3 0.2 Dha 0.0 0.0 0.1 SA 0.0 0.0 0.0 Transition matrix
  7. Learning and playing raga Learning a raga • Feed known

    compositions as input • Using these compute the transition matrix Playing a raga Predict using last note Play predicted note Initial note
  8. Playing a raga: an aside 0.2 0.3 0.5 Re Pa

    Ni probability pdf → cdf Stack up these sticks Ga!
  9. Open questions What really makes up a raga? Does the

    learnt transition matrix capture everything? What are the structures that are hidden? How does one encode rhythm? How does one learn directly from music (and not notation!)? Not bad!
  10. Similar Works • Probabilistic Model: https://link.medium.com/hKWxA7dix0 • Genetic Algorithm :

    An interactive neo-Darwinian algorithm for musical creativity by Dr. Milind Watve