A Novel Technique for Detecting Depressive Disorder: A Speech Database-Based Approach.

Area of Science:

• Psychiatry
• Computational Linguistics
• Machine Learning

Background:

• Depression is a leading global mental health issue, projected to be the most prevalent by 2030.
• Early diagnosis of depressive disorder is crucial for effective treatment and management.
• Automatic Depression Detection (ADD) systems using speech offer a promising avenue for early-stage diagnosis.

Purpose of the Study:

• To develop a novel, labeled audio distress interview database in the Bengali language for depression detection.
• To identify and present a set of hand-crafted acoustic features effective for depression detection using speech signals.
• To validate the utility of the database and acoustic features through a baseline machine learning model.

Main Methods:

• Creation of a unique Bengali speech database comprising audio responses from both depressed and non-depressed individuals.
• Extraction and analysis of hand-crafted acoustic features from speech signals.
• Implementation of a baseline machine learning model to evaluate the predictive efficacy of the developed features and database.

Main Results:

• The study successfully developed and validated a novel Bengali speech database for depression research.
• A set of acoustic features demonstrated effectiveness in predicting depression from speech signals.
• The baseline machine learning model confirmed the quality of the database and the efficacy of the feature set.

Conclusions:

• The developed annotated database is a valuable resource for clinicians and researchers in the field of mental health.
• The findings support the potential of speech-based analysis for automatic depression detection, particularly in the Bengali-speaking population.
• This work paves the way for developing clinical tools for early depression diagnosis through accessible speech data.