Enhanced sine wave speech with an added cue for pitch improves tone perception in Cantonese

Abstract

Cantonese has six lexical tones (four level and two rising) that distinguish otherwise identical syllables. Although it is known that other acoustic cues besides f0 (e.g., voice quality) enter into tone perception, it is unknown whether f0 alone provides a sufficient cue. To test this, modified Sinewave Speech (SWS) replicas of Cantonese sentences were presented to listeners. SWS replaces vocal tract formants with frequency-modulated sinusoids, and has been shown to support perception of phonemic content of speech. However, because formant trajectories alone lack information about f0 and voice quality, standard SWS is unsuitable for studying tone perception (Feng et al., 2012, JASA131/EL133; Rosen & Hui, 2015, JASA138). To create the stimuli, the lowest sinusoid of the SWS replica (representing F1) was replaced with a complex tone constructed using a bandpass with the same center frequency as F1, just wide enough at any timepoint for two harmonics of an independently specified f0 contour. This two-component tone implies a missing fundamental, allowing simultaneous impression of harmonic direction and F1 direction. Modified and unmodified SWS replicas of Cantonese words were presented to native Cantonese speakers to ascertain whether implied f0 improves tone perception; preliminary results support an effect of modified SWS.