From CCRMA Wiki
Revision as of 04:51, 20 March 2010 by Jbhat (Talk | contribs) (iDIA)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

DIA: Do It A Cappella and iDIA


Premise / Idea

  • Very little objective feedback in a capella rehearsal
  • Singers need a way to visualize their performance so they can reflect upon it
  • Need a visual feedback system for a cappella singers that can be used alone or together as an ensemble
  • Learning tool to train the voice and the ears are very helpful
  • Want management tools which a director can use to better monitor performers


  • A large population of amateur a cappella singers who lack the necessary ear training to effectively self-correct
  • Need for an objective method of performance evaluation to eliminate confusion over "who's hearing it right"
  • Desire to explore basic real-time audio information retrieval techniques
  • The appeal of a game that can help you improve!

What is DIA?

  • Above all, DIA is a real-time visualization tool for use by one or more vocal performers
  • A network application allowing multiple singers to use laptops to see personalized visualizations provided by a host computer
  • Vocal training application providing error and deviance visualization and post-performance statistics
  • A music learning tool for people who can't read music
  • Or, for those who just want to have fun!


  • Multi-track MIDI file loaded on start-up, to which voices are to be compared
  • Multiple inputs connect to one host computer (Server/Director)
  • Networked client computers can connect to the Server
  • Clients, however, do almost no processing, but instead display information sent from the server
  • Clients will receive channel-specific data from the server, like pitch
  • Pitch information will be determined by implementing a version of the YIN pitch detection algorithm
  • Clients may also listenfor data from other users
  • Performance error will be calculated with respect to deviation from MIDI arrangement
  • Deviation from both the closest note and the 'correct' midi note are represented both numerically and visually

Testing / Usage

  • The Harmonics, a Stanford a cappella group, will be able to put DIA through comprehensive, rigorous testing (hopefully starting this Winter)
  • System allows for easy addition of features
  • As the program becomes more comprehensive, it will have added value for the user
  • Effectiveness of execution will be judged based on ability to provide accurate and useful visual feedback to the user


  • the DIA Team
 - Charlie Forkish
 - Jay Bhat

Milestones (Revised)

  • DATE 1: 11/15 - accurate and robust implementation of YIN pitch detection algorithm with one input and playback of midi files
  • DATE 2: 12/4 - support for multiple inputs and polished visual feedback
  • DATE 3: 12/10 - distribution to multiple client computers over network and statistics report

Future Work / Addons

  • Brightness Feedback
  • Tonal/Vowel Information
  • More 'musical' display (e.g. notes, accidentals)
  • Playback of MIDI scores
  • Pausing
  • Starting Mid-way, Resuming, Jumping in
  • Pitch Bend
  • More extensive report card


  • Extension of DIA planned as a more single-user oriented model
  • Meant to be able to sing along with any track of any Midi file
  • Ultimately meant for normal people to learn to sing their favorite songs better
  • Performances are scored and stored for later playback
  • Concept of being able to sing along with other people who did the same thing
  • Being able to only hear other people in your own locality
  • Still a work in progress!

Link: http://ccrma.stanford.edu/~jbhat/256b/iDIA.zip