Stuart Rosen
Richard J. Baker
Dept. of Phon. & Linguist., Univ. College London, 4 Stephenson Way, London NW1 2HE, England
Although it is well known that psychoacoustically measured auditory filters broaden with increases in level, it is not yet clear what aspect of the acoustic stimulus controls this nonlinearity---that is, whether the change in bandwidth is controlled primarily by the level of the probe or masker. Therefore, a new method has been developed for fitting filter shapes to notched-noise data in which filter parameters depend explicitly on signal level (probe or masker). By applying this technique to notched-noise data in which both fixed-probe and fixed-masker paradigms have been used at a range of levels, it was possible to show that models in which filter parameters depend on probe level fit the data considerably better than models in which filter parameters depend upon masker level. Furthermore, it was possible to describe the nonlinear changes in auditory filter bandwidth at 2 kHz with only five parameters, allowing the implementation of reasonably realistic computational nonlinear auditory filters whose shape depends on their output. [Work supported primarily by the Medical Research Council (UK), with important contributions from the Hearing Research Trust.]