Testing Acoustic Voice Quality Classification Across Languages and Speech Styles

Abstract

Many studies relate acoustic voice quality measures to perceptual classification. We extend this line of research by training a classifier on a balanced set of perceptually annotated voice quality categories with high inter-rater agreement, and test it on speech samples from a different language and on a different speech style. Annotations were done on continuous speech from different laboratory settings. In Experiment 1, we trained a random forest with Standard Chinese and German recordings labelled as modal, breathy, or glottalized. The model had an accuracy of 78.7% on unseen data from the same sample (most important variables were harmonics-to-noise ratio, cepstral-peak prominence, and H1-A2). This model was then used to classify data from a different language (Icelandic, Experiment 2) and to classify a different speech style (German infant-directed speech (IDS), Experiment 3). Cross-linguistic generalizability was high for Icelandic (78.6% accuracy), but lower for German IDS (71.7% accuracy). Accuracy of recordings of adult-directed speech from the same speakers as in Experiment 3 (77%, Experiment 4) suggests that it is the special speech style of IDS, rather than the recording setting that led to lower performance. Results are discussed in terms of efficiency of coding and generalizability across languages and speech styles.

Voice quality and speaking rate in Icelandic rhetorical questions

“…1 Previous research comparing VQ in German RQs and ISQs showed that given the variability of the experimental material, a perceptual analysis was more robust than acoustic measures (Braun et al 2019). Data from 11 female speakers of the current study was also classified in a random forest model (model included measures of HNR, H1-A2, H1-A3, H1*-A3*, cepstral-peak prominence, glottal to noise excitation ratio, and b1 and b2 measurements); the accuracy of the model was 78.6% overall (Braun et al 2021).…”

“…/pɔr/ in borðar), (ii) the initial, stressed syllable of the subject einhver (/e ͡ in/ in /ˈe ͡ ͡ in.k h vεr/ 'anybody'), (iii) the stressed syllable of the object noun, and (iv) the offset of the sentence. Three types of VQ were perceptually classified: modal (neutral mode of phonation, Laver 1980), breathy (audible friction of the air) and glottalized (low frequency irregular vocal fold vibrations, Braun et al 2021). 1 Speaking rate was operationalized as the number of syllables per second.…”