JadeEast
Headphoneus Supremus
- Joined
- Feb 12, 2007
- Posts
- 2,215
- Likes
- 182
Quote:
The present research has demonstrated that listeners can, with varying degrees of success,
hear different types of smiles in the voices of strangers in the absence of visual cues. Listeners
are very good at discriminating ‘Duchenne Smiles’ from ‘No Smiles’. They can also, to a lesser
degree, successfully discriminate ‘Non-Duchenne Smiles’ from ‘No Smiles’, and ‘Suppressed
Smiles’ from ‘No Smiles’. These findings support previous research findings demonstrating that
smiles can be communicated vocally (De Gelder & Vroomen, 2000; Tartter & Braun, 1994).
Listeners do make mistakes however, and appear to relate certain acoustical
characteristics (possibly: increased pitch, increased intensity, increased F1 and F2 dispersion,
and decreased in F2 and F3 dispersion) with more smiley sounding voices.
Listeners have
(potentially misguided) preconceptions of what smiles should sound like. This possibility raises
serious questions about the use of actors in research. Actors too, like the listeners in the present
study, may have prototypical ideals regarding emotional vocal expressions, and if anything these
ideals are reinforced in society through the use of actors (on television, stage, and radio). Results
obtained from studies utilizing actors should not, therefore, be taken as evidence of how
vocalizations of affect are expressed in everyday usage, but more as illustrative of how vocal
affect expression is represented by society.
Link to a paper on this subject: http://peer.ccsd.cnrs.fr/docs/00/49/91/97/PDF/PEER_stage2_10.1016%252Fj.specom.2007.10.001.pdf
It's seems a bit of a messy thing talking about detecting the smile from a voice. We are interpretation "machines": we are always making interpretations of the world. A machine can't - yet and possibly ever - interpret the meanings of smiles and then understand the social cues. I have no doubt that the difference between a smiling voice and non smiling voice could be correlated to a change in some aspect of the signal. There appears to be some research done with spectral analysis, but the papers lay behind paywalls.
Still, it's the kind of question that lay outside of the concerns of sound reproduction: the differences exists outside of any reproduction technology. You could, just walk up to someone and close your eyes and try to determine if they were smiling. No cables, no driver, no amp, no dac, no technology.
We can probe the voice signal with instruments, but we will not be able to grasp the smiles inside the signal: the smile is a human interpretation. We can make a correlation to some parameter of the sound, but it's not the smile itself.