Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Novel Stochastic Transformer-based Approach for Post-Traumatic Stress Disorder Detection using Audio Recording of Clinical Interviews

Mamadou Dia, Ghazaleh Khodabandelou, Alice Othmani

TL;DR

A deep learning-based approach that achieves state-of-the-art performances for PTSD detection using audio recordings during clinical interviews is proposed, based on MFCC low-level features extracted from audio recordings of clinical interviews, followed by deep high-level learning using a Stochastic Transformer.

Abstract

Post-traumatic stress disorder (PTSD) is a mental disorder that can be developed after witnessing or experiencing extremely traumatic events. PTSD can affect anyone, regardless of ethnicity, or culture. An estimated one in every eleven people will experience PTSD during their lifetime. The Clinician-Administered PTSD Scale (CAPS) and the PTSD Check List for Civilians (PCL-C) interviews are gold standards in the diagnosis of PTSD. These questionnaires can be fooled by the subject's responses. This work proposes a deep learning-based approach that achieves state-of-the-art performances for PTSD detection using audio recordings during clinical interviews. Our approach is based on MFCC low-level features extracted from audio recordings of clinical interviews, followed by deep high-level learning using a Stochastic Transformer. Our proposed approach achieves state-of-the-art performances with an RMSE of 2.92 on the eDAIC dataset thanks to the stochastic depth, stochastic deep learning layers, and stochastic activation function.

A Novel Stochastic Transformer-based Approach for Post-Traumatic Stress Disorder Detection using Audio Recording of Clinical Interviews

TL;DR

A deep learning-based approach that achieves state-of-the-art performances for PTSD detection using audio recordings during clinical interviews is proposed, based on MFCC low-level features extracted from audio recordings of clinical interviews, followed by deep high-level learning using a Stochastic Transformer.

Abstract

Post-traumatic stress disorder (PTSD) is a mental disorder that can be developed after witnessing or experiencing extremely traumatic events. PTSD can affect anyone, regardless of ethnicity, or culture. An estimated one in every eleven people will experience PTSD during their lifetime. The Clinician-Administered PTSD Scale (CAPS) and the PTSD Check List for Civilians (PCL-C) interviews are gold standards in the diagnosis of PTSD. These questionnaires can be fooled by the subject's responses. This work proposes a deep learning-based approach that achieves state-of-the-art performances for PTSD detection using audio recordings during clinical interviews. Our approach is based on MFCC low-level features extracted from audio recordings of clinical interviews, followed by deep high-level learning using a Stochastic Transformer. Our proposed approach achieves state-of-the-art performances with an RMSE of 2.92 on the eDAIC dataset thanks to the stochastic depth, stochastic deep learning layers, and stochastic activation function.
Paper Structure (22 sections, 8 equations, 2 figures, 2 tables)

This paper contains 22 sections, 8 equations, 2 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Post-Traumatic Stress Disorder
  3. Deep Learning
  4. Related work
  5. Study of PTSD
  6. Machine Learning for PTSD detection with audio speech
  7. Deep Learning for PTSD detection using audio speech
  8. Proposed approach
  9. The model
  10. Local-Winner-Take-All layers
  11. Locally Connected layers
  12. Dropout layer
  13. GeLU activation function
  14. Stochastic Depth
  15. Spatial reduction of attention
  16. ...and 7 more sections

Figures (2)

  • Figure 1: Overall architecture of our Stochastic Transformer. The model is composed of 3 main modules: the Patch Creation module (green frame), the Transformer Encoder module (blue frame), and the Regression module (red frame).
  • Figure 2: Scheme describing LCN. Neurons are not connected to all others neurons. LCN are also fed with a patch of the input, instead of the whole input, compared to fully connected layers.