Towards an acoustic-semantic space of extreme metal vocal styles

Description

Extreme vocal techniques typical for metal, e.g. growling or screaming, are characterized by low harmonicity and high roughness and associated with expressive dimensions like "aggressiveness" [Tsai et al., 2010]. Based on audio features, classification into broad style categories is possible [Kalbag & Lerch, 2022]. Which audio features are associated with the perception of emerging expressive techniques/stylistic devices that go beyond previously known categories remains open.

Short phrases were extracted from 105 metal vocal tracks, 10 pilot-rated by subjects for pairwise similarity (45 comparisons). The resulting similarity matrix serves as the basis for a perceptual similarity space computed using multidimensional scaling (MDS). In another pilot experiment, free verbal associations are collected for all 105 song excerpts.

Preliminary analyses reveal a three-dimensional similarity space whose first major axis represents the contrast between harmonic vs. more inharmonic/rough singing (Harmonic-to-Noise Ratio: r = 0.837, p = 0.005; Spectral Complexity: r = -0.959, p < 0.001). While the second perceptual dimension shows no correlations with extracted sound features, the third dimension relates to the position of the higher formants (e.g., F2: r = -0.855, p = 0.003). When projecting the verbal associations into the acoustic space, descriptions such as "demonic," "scratchy," "powerful," or "angelic" are located at specific positions.

Period	18 Mar 2024
Event title	50th DAGA Annual Conference for Acoustics
Event type	Conference
Conference number	50
Location	Hanover, Germany, Lower SaxonyShow on map
Degree of Recognition	International