Spread Lips + Raised Larynx + Raised Pitch = Smiled Speech?
Speaker: Eva Lasarcyk
Institution: Saarland University
Abstract:
When you produce smiled speech, people can "hear" the smile on your face although they might not even see you - e.g. when talking on the phone. The idea of the experiment presented in this talk was bringing together acoustic as well as speech production aspects of smiling - the main question being: How does vocal tract shortening affect the perceived "smileyness" of speech? That is how does a particular articulatory setting influence the acoustic output of speech?
In a synthetic production study, we varied three articulatory parameters independently of each other and tested the relative impact of them in an online listening experiment, thus only providing the acoustic dimension to the participants. For the creation of the audio samples we used an artiulatory speech synthesis system based on a 3D geometric model of the vocal tract.
The results showed that pitch (fundamental frequency) had the highest impact on perceived smileyness. The other two parameters were supposed to account for the vocal tract shortening: lip spreading and larynx raising. It turned out that they would only work well if supported by raised fundamental frequency.
The results also vary with respect to the linguistic (phonemic) form of the stimuli used, showing that the perception of smileyness is depending on the phone that bears the smile: In this sense, /i:/ is inherently "smiled" whereas e.g. /a:/ is not - due to phonemic lip position. A similar interaction occurs with the larynx position.
