I will be in Paris next week to give an invited seminar on the MorpheuS music generation project at l'institut de recherche et coordination accoustique/musique (IRCAM) in Paris. Read the full announcement on Ircam's website

Title: Morphing Music According to a Long-term Tension Profile and Detected Patterns

When: Wednesday 20th April, 2016 at 12h
Where: Ircam, salle Stravinsky

Abstract:
A major problem with much of the automatically generated music is that it lacks a structure and long-term coherence. In this research, we have defined the music generation problem as a combinatorial optimization problem. The advantage of this approach is that it gives us the freedom to impose both hard and soft constraints. In this talk different types of structure, such as repeated, transposed patterns, are constrained. The Cosiatec pattern detection algorithm was used to find maximum translatable patterns. These patterns were then used to constrain the output of a music generation algorithm called MorpheuS. A second form of structure, which is soft constrained, is a tension profile. This type of tension could be relevant to, for instance, automatic generation of game or video music. We have developed a model that captures aspects of tonal tension based on the spiral array (Chew, 2012), a three dimensional model for tonality. Our approach first segments a musical excerpt into equal length subdivisions and maps the notes to clouds of points in the spiral array. Using vector-­based methods, four aspects of tonal tension are identified from these clouds (Herremans & Chew, 2016): the cloud diameter measures the dispersion of clusters of notes in tonal space; the cloud momentum measures the movement of pitch sets in the spiral array; tensile strain measures the distance between the local and global tonal context; finally, angular momentum measures the directional change for movements in tonal space. The results of generating polyphonic piano music with constrained patterns and tension profile show larger structures and sound musically interesting.