You can check these documents.
Making music using concepts from Schenkerian Analysis and Chord Spaces
and Probabilistic Model for Chordal Progressions
But this question is complex, how do you want, for example, to say that the accurate and compact presentation of music signals is a key component of large-scale music-based content applications such as music content management and almost duplicate sound detection. In this case, you are working on a large scale C, which looks like this:
C - D - E - F - G - A - B
at intervals
C - STEP - D - STEP - E - HALF STEP - F - STEP - G - STEP - A - STEP - B - HALF STEP - C -
Now the chord is formed by the distance between notes, for example
C major chord is formed by CEG D minor chord is formed by DFA E minor chord is formed by EGB F major chord is formed by FAC G major chord is formed by GBD A minor chord is formed by ACE B dim chord is formed by BDF
The problem that you describe is not yet resolved, despite numerous studies in this area. So, for example, take a look at other articles where they offer average levels of summing musical signals based on chord progressions. Thus, chord progressions are recognized by musical signals based on a controlled learning model, and recognition accuracy is improved locally by examining the n-best candidates.
So, you can explore the properties of chordal progressions, and then calculate the histogram from the tried chordal progressions as a summary of the musical signal. Then, with a generalization based on chords, harmonic progressions and tonal structures of musical signals can be described.
But how to do that? Well, you need musical data sets (> 70,000 songs?) So you can get the relevant information ...
cMinor
source share