| Popis: |
A framework is proposed for generating musically similar and interesting variationsof a given monophonic melody. In this work, the focus is on rock/pop guitar andbass-guitar melodies with the aim of extensions to other instruments and musicalstyles. The original audio melody in audio format is first segmented into its componentnotes using onset detection and pitch estimation. Clustering is performed onthe pitches to obtain a symbolic representation of pitch sequences in it. The noteonsets are aligned with the estimated meter of the melody for a time-homogeneoussymbolization of the rhythm in terms of onsets/rests and the metrical locations oftheir occurrence. A joint representation based on the cross-product of these twoindividual representations is used to train the prediction framework - the variable lengthMarkov chain. It is hypothesized that such a model will rearrange, whilemaintaining certain coherence, the segments of the original melody. The musicalquality of the generated melodies was evaluated through a questionnaire by a groupof experts and received an overall positive response. |
