Abstract:
Similarities in timbre among 36 broadband complex tone stimuli were measured by a rating-scale method to derive multidimensional perceptual space for the stimuli. Fundamental frequencies of the stimuli were 100, 200, or 400 Hz, spectral envelopes were 0, -3, -6, or -9 dB/oct, and the numbers of components were 20, 30, or 40. The obtained similarities were analyzed by the method of multidimensional scaling, and a three-dimensional configuration was drawn. Distances among stimuli in the configuration did not show good agreement with differences in the physical frequency spectrum. The distances in the perceptual space were then compared with differences in the ``masked frequency spectrum'' [Ozawa et al., Proc. 15th ICA III, 465--468 (1995)] which was the frequency spectrum in terms of the loudness of the narrow-band frequency component of the stimuli partially masked by each other. The agreement between them was better than that in the case of the physical frequency spectrum. This suggests that the masked frequency spectrum of a sound is a good approximation to the final spectrum perceived in hearing.