Extended Abstract: Relationships Between Acoustic Features and Perceptual Segmentation of Music Audio

Abstract

For a successful practical sonification, boundaries where the information state changes need to be made readily apparent perceptually. I describe time series analysis techniques for the detection of segments perceived implicitly during continuous music audio. The detected segments are compared with those measured computationally, or determined musicologically, on the input audio. The degree to which perceptual segmentation can be predicted is discussed, together with some of the factors apparently responsible. This may give useful cues as to how best to structure sonifications for informational purposes.