Limits of Intelligibility of Accelerated Synthesized Speech by Inexperienced Sighted and Experienced Blind Listeners

Abstract

People with blindness who use computer systems cannot use the standard CRT display to access information presented by the computer. The main medium for people with blindness to access computers is through electronic speech synthesizers. One of the characteristics of computer users with blindness is to increase the speech rate of the synthesizer to rates more than double normal conversational speech rates. To the casual listener the increased speech rates seem unintelligible. There have not been quantitative measurements of this observed behavior to determine the limits of speech intelligibility and behaviors at high speech rates. This study is designed to investigate the limits of speech intelligibility of accelerated synthesized speech for both inexperienced and experienced listeners. The results of this experiment show that intelligibility decreases linearly with speech rate. There is no significant difference between the percentage of words correct between the experienced and inexperienced group. There is a significant difference in the number of words confused between the two groups. The lack of group effects in the percentage of words correct is due to a bimodal distribution of scores in the experienced group. While subjects in the experienced group clearly demonstrated the highest intelligibility the group also have some of the lowest intelligibility scores.