Multidimensional Perceptual Unfolding of Spatially Processed Speech I: Deriving Stimulus Space Using INDSCAL
William L. Martens, Nick Zacharov · Journal of the Audio Engineering Society · 2000
Multidimensional perceptual unfolding was executed for a set of spatially-processed speech samples reproduced via headphones. This paper describes only the first stage of a two-part study employing the analytic technique termed external unfolding. In this first stage, global dissimilarity ratings were made for all pairwise comparisons of the experimental stimuli under each of four listening conditions, which included broadband or bandlimited speech samples presented either simultaneously or sequentially. These four datasets were analyzed independently using INDSCAL (INdividual Differences SCALing), a method for processing inter-stimulus dissimilarity data that has specific advantages over classical MultiDimensional Scaling (MDS) analysis. First, it is designed to characterize quantitatively the individual differences in responses obtained from a group of experimental subjects. Second, the spatial configuration of points derived for the experimental stimuli, termed the Stimulus Space, has an inherently unique orientation that has none of the ambiguity that makes the interpretation of classical MDS results problematic. The Stimulus Space derived in the first stage of this study is one of two inputs required for external unfolding. It is combined with discrete attribute ratings collected in the second stage of this study to reveal the principal perceptual attributes of the set of spatially-processed speech samples, and their relative salience under various conditions.