Effect of early echoes and reverberation on the subjective evaluation of speech

Abstract

Using a male and a female speaker, test sentences were created having 25 combinations of early echoes (EEs), i.e., echoes arriving within 10 msec of the direct sound, and four values of reverberation time (RT). These test stimuli were bandlimited from 0.5 to 3.3 kHz and evaluated by subjects in an anchored rating test according to the subject's preference for use in telephony. Analysis of the data by a multidimensional scaling technique showed that preference decreases slightly with increasing RT up to 0.25 sec and more strongly for RTs from 0.25 to 0.5 sec. Addition of six EEs distributed over 7 msec decreases preference three times more than an RT of 0.25 sec and somewhat less than an increase in RT from 0.25 to 0.5 sec. Periodically spaced EEs subjectively degrade speech more than an equal number of irregularly‐spaced ones. Four bunched EEs spread over 0.6 msec are preferred to four EEs distributed over a 4‐msec interval.