Impress: A Machine Learning Approach to Soundscape Affect Classification for a Music Performance Environment

Thorogood, Miles and Pasquier, Philippe

Proceedings of the International Conference on New Interfaces for Musical Expression

Soundscape composition in improvisation and performance contexts involves manyprocesses that can become overwhelming for a performer, impacting on thequality of the composition. One important task is evaluating the mood of acomposition for evoking accurate associations and memories of a soundscape. Anew system that uses supervised machine learning is presented for theacquisition and realtime feedback of soundscape affect. A model of sound-scape mood is created by users entering evaluations of audio environmentsusing a mobile device. The same device then provides feedback to the user ofthe predicted mood of other audio environments. We used a features vector ofTotal Loudness and MFCC extracted from an audio signal to build a multipleregression models. The evaluation of the system shows the tool is effective inpredicting soundscape affect.