Tempo Induction and Beat Tracking for Audio Signals by wku19297


									Tempo Induction
and Beat Tracking
for Audio Signals
MUMT 611, February 2005
Assignment 3
Paul Kolesnik
 Introduction and Definitions
 Overview of Works
 Summary of Common Techniques
 Conclusion

                                 2 of 19
   Concepts
     Tempo
        rate at which a musical piece is played

        score time units per real time unit (eg. beat per minute)

     Beat
        unit in a sequence of impulses which define the tempo of a
         musical piece
        Has no exclusive definition, can be ambiguous and context-
         dependent (eg. 60 bpm vs. 120 bpm)
        Has period (inter-beat interval) and phase (estimated to beat
         position in score)

                                                                     3 of 19
   Tempo Induction
           Process of estimating basic tempo from musical data
   Beat Tracking
     Definition
        The process of extracting beat information from a musical
         score (based on tempo information)
        Differentiated from score following through the absence of a
     Applications
        Performance analysis, perceptual modeling, audio content
         analysis and transcription, performance synchronization

                                                                  4 of 19
   Beat Tracking of Musical Data
        MIDI - symbolic representation, information
         needed for beat tracking is encoded directly in the
         data (eg. note onsets)
        Audio - needs preprocessing to extract symbolic
         representation of the signal

                                                        5 of 19
Overview of Works
   Schloss (1985)
          One of the earliest works
          Onsets detected as peaks in the slope of amplitude envelope of
           a high-pass filtered signal (HFC analysis).
  Allen   and Dannenberg (1991)
          Definition of a period and phase, the concept of beam search
           using multiple tempo / beat hypotheses
          Not clear if MIDI or Audio is used as input
  Goto    and Muraoka (1995-2001)
          Extensive work on beat-tracking systems with and without
           drums, combined in 2001
          Early system (with drums): examines frequencies of snare and
           bass drums, finds onsets and matches them to a set of pre-
           stored drum patterns
                                                                    6 of 19
Overview of Works
  Goto   and Muraoka (ctd.)
       Disadvantage: limited to a specific musical style.
        Advantage: highly successful for this style.
       Later system (without drums): using higher-level musical
        knowledge - chord changes - to determine low-level beat
       Both systems combined into a single system that uses a
        combination of drum indications, chord changes and
        onset indications.
       Systems based on multiple agent architecture, with each
        agent predicting beat times using different strategies.

                                                           7 of 19
Overview of Works
  Goto     and Muraoka (ctd.)
      Musical-knowledge rules used by the system
         Onset-related rules
            (a-1) A frequent inter-onset interval is likely to be an
             inter-beat interval
            (a-2) Onset times tend to coincide with beat times
             (i.e. sounds are likely to occur on beats)
         Chord-related rules

            (b-1) Chords are more likely to change on beat times
             than on other positions

                                                                8 of 19
Overview of Works
  Goto   and Muraoka (ctd.)
       Chord-related rules (ctd)
           (b-2) Chords are more likely to change on half-note
            times than on other positions
           (b-3) Chords are more likely to change at the
            beginnings of measures than at other positions of
       Drum pattern-related rules (ctd)
           (c-1) The beginning of the input drum pattern
            indicates a half-note time
           (c-2) The input drum pattern has an appropriate
            inter-beat interval

                                                            9 of 19
Overview of Works
  Scheirer   (1998)
       Based on tuned resonators
       Signal is split in 6 frequency bands; amplitude envolopes
        are extracted and passed through a set of 150 comb
        filters (each representing possible tempo on a discretized
        scale); output summed across frequency bands; highest
        value determines the tempo and phase
       Problem: filter spacing in relation of tempo

                                                            10 of 19
Overview of Works
  Dixon   (2001)
       Accepts audio or symbolic (MIDI) data
       Two stages of processing

          Tempo induction (examines times between pairs of
          Beat tracking (determines the period / alignment of
           the beats)

                                                          11 of 19
Overview of Works
  Tempo   Induction
      Examines times between pairs of note onsets
      Uses clustering algorithm to determine significant
       clusters of inter-onset intervals
      Each cluster represents a hypothetical tempo

      Output: a list of ranked tempo hypotheses

      For audio, significant preprocessing (onset detection) is
       needed -- this is done using amplitude envelope
       techniques described in (Schloss 1985)

                                                           12 of 19
Overview of Works
  Beat   Tracking
       Tempo induction calculates the frequency/period of the
       Beat tracking calculates the phase
       This is done using the multiple hypothesis search, with
        the best output score determining the identified beat
       Each hypothesis search is conducted by a beat tracking
        agent, which predicts a beat time and matches it to
        rhythmic events, adjusts its hypothesis, creates a new
        one or deletes it if two identical hypotheses are reached

                                                            13 of 19
Overview of Works
  Beat   Tracking
       For each tempo induction-generated hypothesis, there is
        a group of agents created to track the piece at this tempo
       Works based on the assumption that there is at least one
        event in the initial section of music (5 sec) that coincides
        with the beat time
       Agent adjustments, creations, deletions take place based
        on the analyzed information
       Decisions are made based on how evenly spaced events
        are, how often they match the expected beat time values,
        and salience of matched events

                                                             14 of 19
Overview of Works
  Musical   Salience
       How significant any particular event is based on the
        higher-level knowledge of the musical context in which it
        takes place
       Examples of salience factors
           Note duration
           Simultaneous note density
           Amplitude
           Pitch

                                                            15 of 19
Overview of Works
  Davies   and Plumbley (2004)
       A realtime audio beat-tracking system, allowing tempo
        changes and different styles
       Performs onset detection using high frequency content
        and complex domain algorithms
       Tempo induction and beat alignment (phase) estimation
        are similar to techniques implemented in Scheirer (1998)
        and Dixon (2001)
       Uses autocorrelation function to determine the beat
        period, and comb filters for beat phase detection

                                                          16 of 19
Summary of Common Concepts
   Audio Tempo Induction and Beat Tracking is more
    complex than MIDI due to its non-symbolic nature
   Note onset is the most popular element used for
    tempo induction
   Common tempo induction techniques are HFC
    analysis for signals with drums present, and
    complex frequency analysis as a more general tool
   Higher-level musical knowledge rules can facilitate
    the tempo induction process

                                                  17 of 19
Summary of Common Concepts
   Beat tracking of the signal involves determining the
    phase (alignment) of the signal and is done based
    on onset detection data and tempo induction
    (period estimation) data
   Multiple hypotheses techniques (introduced as a
    beam search technique by Allen and Dannenberg
    1991) is commonly used and is superior to
    individual solution approach used in early systems,
    since it allows to recover from encountered tempo
    induction and beat tracking errors

                                                   18 of 19
   HTML Bibliography

   Questions

                                         19 of 19

To top