Noneffect upon Speech Intelligibility of Phase Distortortion Produced by High-Pass and Low-Pass Filters
S. Smith, J. C. R. Licklider · The Journal of the Acoustical Society of America · 1954
Although it has long been the general belief that phase distortion produces no audible effect upon speech, it is easy to set up artificial phase distortions (time delays) that do affect both quality and intelligibility. In order to be sure that the effects of filters result from their selective action upon the amplitudes of the various frequency components, and not to the phase shifts they introduce, it is important, therefore, to have an empirical test. The test was made by playing each speech sample twice through each filter to obtain the effect of two filters in cascade. A recorded word list was played through a filter and re-recorded on a second tape. The second tape was then played through the same filter and presented through earphones to an articulation test crew of five. In half the tests, the tape was always run in the forward direction, and the phase shifts introduced by the filter and tape recorder were doubled. In the other half, the first run through the filter was made with the tape reversed end-for-end, and the phase shifts of the first run canceled those of the second. The two procedures were followed with high-pass and with low-pass commercial electronic filters set at the nominal cut-off frequency of 1800 cps. By cascading units, we obtained cut-off slopes of 48, 96, and 288 decibels per octave. In all, 288 articulation tests were run, each with five listeners. There was no difference at all attributable to phase distortion.