Voice Quality after Using Speech Recognition Software: Perceptual Results and Reliability

This study investigates the influence of using speech recognition software on voice quality. Two different groups of speakers (one group of subjects with a heavy daily vocal load and one control group) were subjected to different speech recognition tasks for 2 hours (either using discrete or continuous speech recognition software). Five listeners assessed the voice quality (14 parameters) before and after the task. Reliability of the ratings was evaluated for different types of speech fragments (3 different vowels, a read passage and conversation), for individual perceptual parameters and for male and female voices.