Perception of speaker characteristics with long and short samples

In this article, the perception of speaker’s characteristics is compared using long and short samples. The sound material was an Italian read sentence (seicento-cinquantotto) and the last two syllables ‘‘to’’ from this sentence. Twenty healthy Italian native speakers (10 females, 10 males) read the sentence, with the same recording conditions. In the first experiment, the listeners performed a free categorization of the syllables and then of the sentences. Then, the listeners were asked to give qualifiers to describe the classes they had made. The aim was to obtain the main criteria for the description of long samples (sentences) and short samples (isolated syllables). For the sentences, the gender and prosody criteria were more frequently used by listeners, whereas the pitch and voice quality were preferred to classify the syllables. In the second experiment, listeners rated the same samples as previously, along predefined axes (age, gender, breath, liveliness, etc.). The ratings were compared with acoustic measures. The results show that the ratings for the whole sentence, and for the extracted syllables, can be significantly different, leading to the conclusion that, within a sentence, there are significant local variations of the perceived voice quality and speaker characteristics as gender and age.