Learning effects for two different speech synthesizers with different speech material
Rory A. DePaolis · The Journal of the Acoustical Society of America · 1988
This study examined the relationship between the learning effects due to two inexpensive methods of synthesizing speech, formant synthesis (FS) and linear predictive coding (LPC). An experiment using two groups of six subjects listening to either FS or LPC for four consecutive days was conducted. The test material consisted of contextual and nonsense sentences with written clues, contextual sentences without written clues, and the rhyme test. A prose passage was also presented to the subjects. There was no significant difference between the learning curves for the two synthesizers but there was an overall dramatic learning effect, most noticeably from the first to third presentation across all types of speech material. The synthesizers exhibited equality of mean percent correct word scores for all material over the four sessions, excluding day 1, using the nonsense sentences in which the FS device was more understandable. These results are presented graphically with comparisons to other studies using natural and synthetic speech and predictions on the performance of the two synthesizers in noise.