Speech-based Automatic Recognition Technology for Major Depression Disorder

EasyChair Preprint 1190

Abstract

Depression is one of the common mental illnesses nowadays. It can greatly harm the physical and mental health of patients and cause huge losses to individuals, families and society. Because of the lack of hardware and social prejudice against depression, there are a large number of misdiagnosis and missed diagnosis in hospitals. It is necessary to find an objective and efficient way to help the identification of depression. Previous studies have demonstrated the potential value of speech in this area. The model based on speech can distinguish patients from normal people to a great extent. On this basis, we hope to further predict the severity of depression through speech. In this paper, a total of 240 subjects were recruited to participate in the experiment. Their depression scores were measured using the PHQ9 scale, and their corresponding speech data were recorded under the self-introduction situation. Then, the effective voice features were extracted and the PCA was conducted for feature dimensionality reduction. Finally, utilizing several classical machine learning method, the depression degree classification models were constructed. This study is an attempt of the interdisciplinary study of psychology and computer science. It is hoped that it will provide new ideas for the related work of mental health monitoring.

Keyphrases: Classification, Depression, machine learning, prediction, speech

@booklet{EasyChair:1190,
  author    = {Zhixin Yang and Hualiang Li and Li Li and Kai Zhang and Chaolin Xiong and Yuzhong Liu},
  title     = {Speech-based Automatic Recognition Technology for Major Depression Disorder},
  howpublished = {EasyChair Preprint 1190},
  year      = {EasyChair, 2019}}