A spectral-envelope synthesis model to study perceptual blend between wind instruments
Wind instrument sounds can be shown to be characterized by pitch-invariant spectral maxima or formants. An acoustical signal-analysis approach is pursued to obtain spectral-envelope descriptions that reveal these pitch- invariant spectral traits. Spectral envelopes are estimated empirically by applying a curve-fitting procedure to a composite distribution of partial-tone frequencies and amplitudes obtained across an instrument’s pitch range. A source-filter synthesis model is designed based on two independent formant filters with their frequency responses matched to the spectral envelope estimates. This is then used in perceptual experiments in which parameter variations of the synthesis filter are manipulated systematically to investigate their contribution to the degree of per- ceived blend between the synthesized sound and a recorded instrument sound. The perceptual relevance is assessed through two tasks in which participants either produce the best attainable blend by directly controlling synthesis parameters or rate the degree of blend for 5 parameter presets. Behavioral data from both experiments suggest the utility of this formant-based model for correlating pitch-invariant acoustical description with perceptual relevance, as both formant frequency and magnitude appear to affect perceived blend.
Citation : Lembke, S.-A. and McAdams, S. (2012) A spectral-envelope synthesis model to study perceptual blend between wind instruments. Proc. 11th Congrès Français d’Acoustique / IOA Annual Meeting / Acoustics 2012. Nantes, France, pp. 1025–1030
Research Group : Music, Technology and Innovation Research Centre
Research Institute : Music, Technology and Innovation - Institute for Sonic Creativity (MTI2)
Peer Reviewed : No
- Leicester Media School