Comprehension of synthesized speech while driving and in the lab

Abstract

Two studies were conducted to measure the comprehensibility of synthetic speech with current text-tospeech technology. Baseline measurements for each subject were obtained using recorded natural speech. The first study was conducted in a quiet lab with no distractions. Half the subjects were allowed to take notes while listening, the other half were not. Findings showed that there was no significant difference in comprehension of synthetic speech among the five different text-to-speech engines used. Subjects that did not take notes performed significantly worse for all synthetic voice tasks when compared to natural speech tasks. In the second study, in addition to measuring comprehension levels, driving performance was examined. Findings suggest that under low workload conditions (very predictable two-lane roads with minimal traffic, no intersections and driving with cruise control), listening to messages does not interfere with driving. This holds for both natural and text-to-speech presentations, though comprehension of natural speech is better and subjects felt so.