Using Deep Neural Networks for Detecting Depression from Speech

Diagnosing depression has become a major concern in the last years that led the research community to try innovative and creative ways of recognizing it. This paper proposes a system that can identify depression from speech samples based on a classification process performed by deep neural networks....

Full description

Saved in:
Bibliographic Details
Published in2023 31st European Signal Processing Conference (EUSIPCO) pp. 411 - 415
Main Authors Gheorghe, Mirela, Mihalache, Serban, Burileanu, Dragos
Format Conference Proceeding
LanguageEnglish
Published EURASIP 04.09.2023
Subjects
Artificial neural networks
DAIC-WOZ
deep neural networks
Depression
depression detection
Europe
Interviews
MODMA
Signal processing
Speech processing
speech technology
Online AccessGet full text

Cover

More Information
Summary:Diagnosing depression has become a major concern in the last years that led the research community to try innovative and creative ways of recognizing it. This paper proposes a system that can identify depression from speech samples based on a classification process performed by deep neural networks. The system was tested on the Distress Analysis Interview Corpus of human and computer interviews (DAIC-WOZ) and the Multi-modal Open Dataset for Mental-disorder Analysis (MODMA), two spontaneous speech databases recorded in realistic conditions and in two different languages. The system is able to generalize well on previously unseen data. Improvements have been obtained over other results reported in literature, yielding an unweighted accuracy (UA) of 91.25% and a weighted accuracy (WA) of 92.10% for DAIC-WOZ.
ISSN:2076-1465
DOI:10.23919/EUSIPCO58844.2023.10289973