The effects of breath sounds on the perception of synthetic speech

When preparing to speak, talkers typically take a breath. The perceptual effect of adding naturally produced breath intake sounds to synthetic speech was examined. In experiment 1, subjects were better at transcribing synthesized sentences that were preceded by a breath sound than those that were not, in addition to the improvement due to practice that is typically found with synthetic speech. Experiment 2 found that replacing the breath with the spectrally similar sound of rustling leaves had no effect on the accuracy. Experiment 3 had breaths before randomly selected sentences. Only the practice effect was significant, though there was a tendency for sentences with the breath sounds to be remembered better. In experiment 4, we tested whether the appropriateness of the breath sound to the sentence size (relatively short or long) affected the use of the breath sound. Appropriateness had no effect, perhaps because the range of sentence durations was too small. Experiment 5 replicated experiment 1 but used leaf sounds rather than silence in the nonbreath sentences. The presence of breath was again found to aid recall. Overall, the current results indicate that adding the breath intake sound to synthetic sentences improves listeners' ability to recall those sentences.