ASA 130th Meeting - St. Louis, MO - 1995 Nov 27 .. Dec 01

3aPP3. Determinants of the perceptual similarity of complex filter shapes.

Yijian Zheng

Charles S. Watson

Dept. of Speech and Hearing Sciences, Indiana Univ., Bloomington, IN 47405

Sounds were created by passing an increasing-frequency 300-ms sawtooth (120--170 Hz) through each of the 15 complex filters. The filters were created by varying two parameters of a pair of overlapping second-order filters (CF: 500 and 1500 Hz). The parameters varied were the width of the upper filter (Q: 1, 3, 8) and the relative amplitudes of the two filter peaks (+12, +6, 0, -6, or -12 dB). Listeners judged the similarity of pairs of these sounds, equated for energy, using a ten-point scale. A multidimensional scaling (MDS) analysis suggested that when Q of the upper filter was low, the sounds were distinguished entirely on the basis of the relative amplitudes of the filters. For high upper-filter Q values, a different intensity-related dimension was the basis for distinguishing among the sounds, which were essentially identical on dimension No. 1. The two dimensions are associated with overall pitch, and with the salience of the high-frequency spectral peak. Various physical models are fitted to the data. [Work supported by NIDCD and by AFOSR.]