MUSICAL STYLE PERCEPTION BY A LINEAR AUTO-ASSOCIATOR MODEL AND HUMAN LISTENERS

The present research adapts to musical style perception a simulation approach previously used with visual stimuli (i.e., faces), which can handle a variety of tasks such as identification, recognition and categorization. A linear auto-associator was trained with musical excerpts of three composers (Bach, Mozart, and Beethoven). Based on its learned representation, a linear classifier (i.e., Adaline) was trained and its ability to categorize new excerpts from these three composers was evaluated. A subset of the excerpts used to train the linear models is currently tested in free and constrained classification tasks with human listeners.