The increase in cases of depression in recent years makes necessary to develop new objective and reliable tools and instruments for the early identification of this condition. In this context, one of the most efficacious and efficient approaches seems to be the acoustic analysis of the speech characteristics of depressed patients. The present study analyzes how depressive states affects the spontaneous speech of 68 depressed patients of two different languages (Italian and English), divided into 3 groups (English, Italians with mild/moderate severity and Italians with severe severity). We consider 21 acoustic features, all of them good and usual indicators of emotional states in speech, obtained from 5 acoustic low-level descriptors (LLD, i.e. Fundamental frequency, Jitter, Shimmer, Voice breaks, and Intensity). The main results showed, in a novel way, that fundamental frequency and number of pauses allowed to discriminate between English and Italian patients, while no significant differences were observed regarding the severity degree of the depressive symptoms. The aim of this study, corroborated by the results obtained, is to highlight the need to take into consideration the language variable when developing an automated speech analysis for detecting depressive states.

The Role of Language in Building Automatic Models for Depression Detection

Raimo G.;Conson M.;Amorese T.;Cuciniello M.;Greco C.;Cordasco G.;Esposito A.
2022

Abstract

The increase in cases of depression in recent years makes necessary to develop new objective and reliable tools and instruments for the early identification of this condition. In this context, one of the most efficacious and efficient approaches seems to be the acoustic analysis of the speech characteristics of depressed patients. The present study analyzes how depressive states affects the spontaneous speech of 68 depressed patients of two different languages (Italian and English), divided into 3 groups (English, Italians with mild/moderate severity and Italians with severe severity). We consider 21 acoustic features, all of them good and usual indicators of emotional states in speech, obtained from 5 acoustic low-level descriptors (LLD, i.e. Fundamental frequency, Jitter, Shimmer, Voice breaks, and Intensity). The main results showed, in a novel way, that fundamental frequency and number of pauses allowed to discriminate between English and Italian patients, while no significant differences were observed regarding the severity degree of the depressive symptoms. The aim of this study, corroborated by the results obtained, is to highlight the need to take into consideration the language variable when developing an automated speech analysis for detecting depressive states.
978-1-6654-6297-6
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11591/486109
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? ND
social impact