The effects of lexical content, acoustic and linguistic variability, and vocoding on voice cue perception

Abstract

Perceptual differences in voice cues, such as fundamental frequency (F0) and vocal tract length (VTL), can facilitate speech understanding in challenging conditions. Yet, we hypothesized that in the presence of spectrotemporal signal degradations, as imposed by cochlear implants (CIs) and vocoders, acoustic cues that overlap for voice perception and phonemic categorization could be mistaken for one another, leading to a strong interaction between linguistic and indexical (talker-specific) content. Fifteen normal-hearing participants performed an odd-one-out adaptive task measuring just-noticeable differences (JNDs) in F0 and VTL. Items used were words (lexical content) or time-reversed words (no lexical content). The use of lexical content was either promoted (by using variable items across comparison intervals) or not (fixed item). Finally, stimuli were presented without or with vocoding. Results showed that JNDs for both F0 and VTL were significantly smaller (better) for non-vocoded compared with vocoded speech and for fixed compared with variable items. Lexical content (forward vs reversed) affected VTL JNDs in the variable item condition, but F0 JNDs only in the non-vocoded, fixed condition. In conclusion, lexical content had a positive top–down effect on VTL perception when acoustic and linguistic variability was present but not on F0 perception. Lexical advantage persisted in the most degraded conditions and vocoding even enhanced the effect of item variability, suggesting that linguistic content could support compensation for poor voice perception in CI users.

Top-down effect of dialogue coherence on perceived speaker identity

“…The acoustic features of signal may therefore not have been identical across Experiments 1 and 2. However, the acoustic voice cues that have been shown to be most salient for the discrimination between voices of different talkers—fundamental frequency (F0) 39 , 44 – 46 and vocal tract length 47 —were not altered by Experiment 2. The long term average spectrum 48 , another cue used by listeners to differentiate between voices 49 , also remained consistent across our experiments.…”

“…Though our results strongly implicate pragmatic context in speaker change detection, our experimental design does not allow us so strictly rule out any influence of acoustic cues. Research suggests a strong interaction between top-down lexical and sentential content and indexical (talker-specific) content 39 , 40 : the presence of linguistic context at the word and sentence levels affects participants’ discrimination of voice cues. In order to test the causal role of pragmatic context in speaker change detection and to further rule out any effect of acoustic cues in Experiment 1, we ran a follow up experiment in which we removed all pragmatic information from the same stimuli, leaving acoustic information intact.…”

“…Research shows top-down effects of linguistic content on the differentiation of voice cues that are relevant to speaker differentiation. The presence of linguistic content at the word level, for example, increases participants’ ability to differentiate between vocal tract lengths of voices 39 , and both vocal tract length and F0 differentiation increases in the context of sentences 50 . These findings indicate a top-down effect of linguistic content on how we integrate acoustic information in speaker voice discrimination processes, meaning that participants in our experiment may have actually been more sensitive to differences in acoustic features in Experiment 1 compared to Experiment 2.…”

Lexical effects on the perceived clarity of noise-vocoded speech in younger and older listeners

“…In the current study, an 8-channel vocoder was used to simulate the same number of electrode contact points in each vocoder condition, but spectral resolution was varied by manipulating the sharpness of the bandpass filter slopes to simulate low, medium, and high spread of excitation in the cochlea. Decreased spectral resolution via simulation of increased channel interaction has been found to result in less accurate speech recognition (Fu and Nogaki, 2005;Bingabr et al, 2008;Wang et al, 2012;Oxenham and Kreft, 2014;Winn et al, 2016;Mehta et al, 2020), less accurate pitch perception (Crew et al, 2012;Mehta and Oxenham, 2017), increased listening effort (Winn et al, 2016), and limitations in the perception of non-linguistic aspects of speech, such as voice cue perception (Gaudrain and Başkent, 2015;Koelewijn et al, 2021). In the current study, while there was a main effect of vocoder condition, only differences between the condition with the worst spectral resolution (HS; 4th order, 24 dB/octave) and the conditions with increasingly more favorable spectral resolutions (MS and LS; 8th order, 48 dB/octave and 12th order, 72 dB/octave, respectively) emerged.…”

“…To investigate the effect of spectral resolution on speech clarity, the current study manipulated the sharpness of the slope of the bandpass filters to simulate current spread in the cochlea. The amount of spread of excitation in the cochlea determines the extent to which individual stimulation channels of the implant interact (e.g., Black and Clark, 1980;Bingabr et al, 2008;Gaudrain and Başkent, 2015;Koelewijn et al, 2021). Three vocoder conditions were included to simulate low spread (LS), medium spread (MS), and high spread (HS) of excitation (and decreasing spectral resolution, respectively), in order to obtain varying degrees of degradation.…”