Creating musical structure from the temporal dynamics of soundscapes

Acoustic signal mixtures from the environment often manifest interesting structural properties. These features can be used to generate musical forms, but access is a problem. Here, Probabilistic Latent Component Analysis (PLCA) has been used to automatically extract independent sources from a mixed signal of environmental sounds. Excitation information corresponding to the temporal evolution of independent acoustic sources in the original environment is extracted. Given this information, we map the temporal dynamics of each independent sound source to a variety of parameters in a musical composition. Thus, the temporal flow of auditory events in an environmental soundscape is used to determine musical structure. After a discussion of PLCA, related composers and music compositions are highlighted. Thereafter, an approach to mapping the temporal dynamics of a given soundscape to a musical setting using source separation techniques is illustrated.