An investigation of linguistic stress and articulatory vowel characteristics for automatic depression classification. (January 2019)

• Record Type: Journal Article
• Authors: Stasak, Brian; Epps, Julien; Goecke, Roland
• Is Part Of: Computer speech & language. Volume 53(2019)
• Page Start: 140
• Page End: 155
• Publication Date: 2019-01
• Subjects: Hypoarticulation -- Paralinguistics -- Psychomotor retardation -- Vowel quadrilateral; Speech processing systems -- Periodicals; Automatic speech recognition -- Periodicals; Computers -- Periodicals; Linguistics -- Periodicals; Speech-Language Pathology -- Periodicals; Traitement automatique de la parole -- Périodiques; Reconnaissance automatique de la parole -- Périodiques; Automatic speech recognition; Speech processing systems; Electronic journals; Periodicals; 006.454
• DOI: 10.1016/j.csl.2018.08.001
• Languages: English
• ISSNs: 0885-2308

Abstract

Highlights: Investigation of depression discrimination via vowel sets based on articulatory parameters. Linguistic stress components found statistically significant differences between non-depressed and clinically depressed speakers. A novel compact linguistic stress feature set produces depression classification improvements over baseline methods. Findings also suggest linguistic stress approach for eliciting speech to improve automatic depression assessment. Abstract: The effects of psychomotor retardation associated with clinical depression are linked to a reduction in variability in acoustic parameters. However, linguistic stress differences between non-depressed and clinically depressed individuals have yet to be investigated. In this paper, by examining vowel articulatory parameters, statistically significant differences in articulatory characteristics are found at a paraphonetic level. For articulatory characteristic features, tongue height and advancement in terms of 'mid' and 'front' vowel sets show similar depression classification performance trends for both the DAIC-WOZ (English) and AViD (German) databases. Considering linguistic stress feature components, for both databases, depressed speakers exhibit shorter vowel durations and less variance for 'low', 'back', and 'rounded' vowel positions. Results for the DAIC-WOZ and AViD datasets using a small set of linguistic stress based features derived from multiple vowel articulatory parameter sets show absolute, statistically significant, gains of 7% and 20% in two-class depression classification performance over baseline approaches. Linguistic stress feature results indicate that specific vowel set analysis provides better discrimination of clinically depressed and non-depressed speakers. Knowledge gleaned from this research allows the design of more effective automatic depression disorder classification systems.