Uni- and multidimensional identification of rise time, spectral slope, and cutoff frequency

Journal/Book: J Acoust Soc Am. 1994; 96: 1380-7.

Abstract: Absolute identification of rise time, spectral slope, and cutoff frequency was evaluated using complex tones comprising the first 40 harmonics of 200 Hz. Ranges of each dimension were restricted to those with relevance to natural music and speech. Baseline unidimensional fixed-background tasks yielded 1.5-2.0 bits of information, indicating that listeners could consistently identify at least three to four categories per dimension. Fixed-and roved-background information transfers implied rise time was perceptually separable from the other dimensions. Spectral slope and cutoff frequency displayed substantial integrality (ratios of fixed-background to roved-background information transfers of 1.08-5.03), but the extent decreased with experience (retest ratios of 1.03-2.64). Stimulus combinations that maximized differences in high-frequency spectral regions were associated with increased information transfer and greater separability, suggesting a functional role for "sharpness" in timbre identifications. Total multidimensional information transfer was closely approximated by the sum of the corresponding unidimensional roved-background values, supporting both the general additivity of information transfer and the feasibility of using unidimensional roved-background tasks to estimate multidimensional resolution efficiently.

Note: Department of Speech and Hearing Sciences University of Washington Seattle 98195.

Keyword(s): Music|. Speech Perception|

Zurück | Weiter