Effects of low-pass filtering on the judgment of vocal affect in speech directed to infants, adults and foreigners

  • Authors:
  • Monja A. Knoll;Maria Uther;Alan Costall

  • Affiliations:
  • Department of Psychology, University of Portsmouth, King Henry Building, King Henry I Street, Portsmouth, PO1 2DY, UK;Centre for Cognition and NeuroImaging, School of Social Sciences, Brunel University, West London, Uxbridge, Middlesex, UB8 3PH, UK;Department of Psychology, University of Portsmouth, King Henry Building, King Henry I Street, Portsmouth, PO1 2DY, UK

  • Venue:
  • Speech Communication
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

Low-pass filtering has been used in emotional research to remove the semantic content from speech on the assumption that the relevant acoustic cues for vocal affect remain intact. This method has also been adapted by recent investigations into the function of infant-directed speech (IDS). Similar to other emotion-related studies that have utilised various levels of low-pass filtering, these IDS investigations have used different frequency cut-offs. However, the effects of applying these different low-pass filters to speech samples on perceptual ratings of vocal affect are not well understood. Samples of natural IDS, foreigner- (FDS) and British adult-directed (ADS) speech were low-pass filtered at four different cut-offs (1200, 1000, 700, and 400Hz), and affective ratings of these were compared to those of the original samples. The samples were also analyzed for mean fundamental frequency (F"0) and F"0 range. Whilst IDS received consistently higher affective ratings for all filters, the results of the adult conditions were more complex. ADS received significantly higher ratings of positive vocal affect than FDS with the lower cut-offs (1000-400Hz), whereas no significant difference between the adult conditions was found in the original and 1200Hz conditions. No difference between the adult conditions was found for encouragement of attention. These findings show that low-pass filtering leaves sufficient vocal affect for detection by raters between IDS and the adult conditions, but that residual semantic information in filters above 1000Hz may have a confounding affect on raters' perception.