Effects of low-pass and high-pass filtering on the identification of environmental sounds

Abstract

Listeners’ abilities to identify environmental sounds under high-pass and low-pass filtering was determined using a constant-stimulus method. The goal was to determine the degree to which listeners rely on information in various frequency regions when attempting to identify naturally occurring sounds. In experiment 1, 70 environmental sounds were low- and high-pass filtered with cutoff values ranging from 300 to 4800 Hz in octave steps. Subjects identified the sounds after an initial familiarization stage in which all sounds were presented without filtering. With low-pass filtering, performance improved from 57% to nearly 100% correct as the cutoff frequency increased. Performance with high-pass filtering was better, with percent correct identification above 70% for all cutoff values. Importance functions derived from these results were used to develop articulation index-like predictions for the identification of a representative set of environmental sounds. [Work supported by NIH.]