Rhythmic Transcription of MIDI Signals

1
Rhythmic Transcription of MIDI Signals

• Carmine Casciato
• MUMT 611
• Thursday, February 10, 2005

2
Uses of Rhythmic Transcription

• Automatic scoring
• Improvisation
• Score following
• Triggering of audio/visual components
• Performance
• Audio classification and retrieval
• Genre classification
• Ethnomusicology considerations
• Sample database management

3
MIDI Signals

• Unidirectional message stream at 3.125KHz
• System Real Time Messages provide Timing Tick
  message
• A simplification of acoustic signals
• No noise, masking effects
• Easily retrieve note onsets, offsets, velocities,
  pitches
• However, no knowledge of acoustic properties of
  sound

4
Difficulties in Rhythmic Transcription

• Expressive performance vs mechanical performance
• Inexact performance of notes
• Syncopations
• Silences
• Grace notes
• Robustness of beat tracker
• Can the tracker recover from incorrect beat
  induction?
• Real time implementation
• (Dixon 2001)

5
Human Limits of Rhythmic Perception

• Two note onsets are deemed synchronous when
  played within 40ms of each other, 70 ms for gt two
  notes
• Piano and orchestral performances exhibit note
  onset asynchronicity of 30-50ms
• Note onset differences of 50ms to 2s give
  rhythmic information
• (Dixon 2001)

6
Evaluation Criteria for Beat Trackers

• Informally - click track of reported beats added
  to signal
• Visually marking the reporting beats
• Comparing reported vs known, correct beats
• (Dixon 2001)

7
Definitions

• Beat - perceived pulses which are approximately
  equally spaced and define the rate at which notes
  in a piece are played
• meterical, score , performance level
• tempo - beats per minute
• Inter-onset Intervals (IOI) - time intervals
  between note onsets
• (Dixon 2001)

8
Approaches - Probabilistic Frameworks

• Cemgil et al (2000) - Bayesian framework, using a
  tempogram (wavelet) and a 10th order Kalman
  Filter to estimate tempo, which is a hidden state
  variable
• Takeda et al (2002) - Hidden Markov models for
  fluctuating note lengths and note sequences,
  estimating both rhythms and tempo
• Raphael (2002) - tempo and rhythm

9
Approaches - Oscillators

• Period and phase that adjusts itself to
  synchronize to IOI input
• Dannenberg and Allen (1990) - weighted IOIs and
  credibility evaluation based on past input
• Meudic (2002) - real time implementation of Dixon
• Induce several beats and attempt to propagate
  them through the signal (agents), then choose the
  best
• Pardo (2004) - Oscillator, compared to Cemgil
  using same corpus

10
Pardo 2004 - Oscillatory Design

• Is a Kalman Filter (Cemgil) or oscillator better
  for online tempo tracking?
• Performance as time series of weights, W, over T
  time steps
• Weight of time step with no note onsets 0,
  increased proportional to of note onsets
• 100ms is minimum IOI allowed, minimum beat period

11
Pardo 2004

• Uses weighted average of last 20 beat periods,
  with one parameter varying degrees of smoothing
• A correction parameter varies how far the period
  and phase of the next predicted beat is changed
  according to known information
• A window size parameter affects how many periods
  may affect the current prediction
• Chose 5000 random values of these three
  parameters, ran each triplet on 99 performances
  of Cemgil corpora

12
Cemgil MIDI/Piano Corpora

• Four pro jazz, four pro classical, three amateur
  piano players
• Yesterday and Michelle, fast, slow and normal,
  captured on a Yamaha Diskclavier
• Available at www.nici.kun.nl/mmm/

13
Pardo 2004 - Error Measurement
(Pardo 2004)

• After finding best parameters values for
  Michelle corpus,
• applied same values to analysis of Yesterday
  corpus
• Compared to Cemgil using that papers defined
  error
• metric, which takes into account both phase and
  period
• errors, to come up with a score

14
Comparison of Approaches
(Pardo 2004)

• Oscillator somewhat better than tempogram alone,
• Somewhat worse than tempogram plus Kalman,
• yet fall within standard deviation (bracketed
  numbers)
• of Kalman scores

15
Other Considerations

• Stylistic information
• Training of tracker
• Musical importance of note
• Duration
• Pitch
• Velocity

16
Bibliography

• Allen, P., and R. Dannenberg. 1990. Tracking
  musical beats in real time. In Proceedings of
  the International Computer Music Conference 1990
  1403.
• Dixon, S. 2001. Automatic extraction of tempo and
  beat from expressive performances. Journal of
  New Music Research 30 (1) 3958.
• Meudic, B. 2002. A causal algorithm for
  beat-tracking. In Proceedings of Conference on
  Understanding and Creating Music.
• Pardo, B. 2004. Tempo tracking with a single
  oscillator. In Proceedings of the International
  Conference on Music Information Retrieval 2004.
• Raphael, C. 2002. A hybrid graphical model for
  rhythmic parsing. Artificial Intelligence 137
  21738.
• Takeda, H., T. Nishimoto, and S. Sagayama. 2002.
  Automatic rhythm transcription from multiphonic
  MIDI signals. In Proceedings of the International
  Conference on Music Information Retrieval 2003.