ASA 125th Meeting Ottawa 1993 May

4pPP8. Profile analysis in the spectro-temporal domain.

T. Houtgast

Adelbert W. Bronkhorst

TNO Inst. for Perception, P. O. Box 23, 3769 ZG Soesterberg, The Netherlands

In this study on energy profile discrimination, the basic sound consists of five Gaussian-shaped tone pulses (wavelets). The choice of the timing and the frequencies of the individual wavelets is such that they are always arranged in a linear array in the log(f)-t spectrogram, with the orientation of the array as an important parameter (horizontal, vertical or any oblique orientation). An additional parameter of these five-wavelets arrays is the mutual separation between the wavelets. For a given orientation and separation, the reference sound contains all five wavelets with equal energy, while in the target sound the (relative) energy of the middle wavelet is reduced: The just-detectable decrement of the middle wavelet is determined with a standard adaptive 2AFC procedure, including a roving-level paradigm. This form of profile analysis with the present five-wavelets arrays shows that (1) the performance is better with a purely temporal pattern (horizontal array) than with a purely spectral pattern (vertical array), and (2) for some specific oblique orientations of the array the performance is extremely poor, with just-detectable decrements of the middle wavelet of the order of 20 dB.