Characterization of Human Emotions and Preferences for Text-to-Speech Systems Using Multimodal Neuroimaging Methods

Abstract

We investigate how listeners’ emotional responses and preferences to text-to-speech (TTS) voices can be quantified using a multimodal approach. Behavioral ratings are combined with EEG- and peripheral-physiology–based features while participants listen to TTS variants differing in prosody and signal quality. Results show reliable links between frontal EEG markers, arousal/valence reports, and user preference, suggesting neurophysiological measures as complementary indicators for TTS evaluation.

BibTeX

@inproceedings{voigtantons2014c25,
  author    = {Rehman Laghari, K. and Gupta, R. and Arndt, S. and Antons, J.-N. and Möller, S. and Falk, T. H.},
  title     = {Characterization of Human Emotions and Preferences for Text-to-Speech Systems Using Multimodal Neuroimaging Methods},
  year      = {2014},
  booktitle = {Paper presented at the IEEE Canadian Conference on Electrical and Computer Engineering (CCECE 2014), Toronto, ON, Canada},
  doi       = {10.1109/CCECE.2014.6901142},
}