Automatic Rhythm Transcription from Multiphonic MIDI Signals

Haruto Takeda    Takuya Nishimoto    Shigeki Sagayama

Graduate School of Information Science and Technology, The University of Tokyo

Hongo, Bunkyo-ku, Tokyo 113-8656 Japan  /   {takeda,nishi,sagayama}@hil.t.u-tokyo.ac.jp

概要:

For rhythm recognition human-performed MIDI signals of polyphonic music, we proposed an approach based on probabilistic models, in which Viterbi search in HMM is used to estimate rhythm and EM Algorithm to estimate tempo. The results of experimental evaluation are also reported.