Effects of Selective Narrow-Band Filtering on the Perception by Normal Listeners of Harvard PB-50 Word Lists
William E. Castle · The Journal of the Acoustical Society of America · 1964
Productions by one adult male speaker of randomized versions of Harvard PB-50 word lists 1 through 12 were stored on tape. Three listening groups, with five listeners each, were asked to respond to the word lists as each list was subjected to a different condition of narrow-band filtering. Groups 1, 2, and 3 always listened under conditions in which the filters had a geometric center frequency of 1080, 1560, and 1920 cps, respectively. The conditions for a particular group were altered from word list to word list by expanding the band limits and/or by using 1, 2, 3, or 4 filters in series. Word-intelligibility scores, expressed as percentages, were found to be demonstrably high for the narrowest frequency bandwidths used. With a single filter, which attenuated signals at the rate of 30 dB or more per octave, scores of 97%–99% were obtained with bandwidths of 240 cps centered at 1560 and 1920 cps and of 480 cps centered at 1080 cps. Regardless of center frequencies, bands as small as from 1200 to 1440 cps wide consistently yielded word-intelligibility scores from 83%–95%, whether 2, 3, or 4 filters were cascaded.