Facial geometry and speech analysis for depression detection.

Area of Science:

• Computer Science
• Psychiatry
• Machine Learning

Background:

• Depression is a widespread mental health disorder with significant global impact.
• Accurate and early detection of depression is crucial for effective treatment.
• Non-verbal cues from facial expressions and speech offer potential indicators of depressive symptomatology.

Purpose of the Study:

• To propose and evaluate a decision support system for depression detection.
• To utilize novel features from facial expression geometry and speech for identifying depression.
• To assess system performance in gender-independent and gender-based modes with various fusion techniques.

Main Methods:

• Extraction of novel features from facial expression geometry and speech.
• Development of a system for interpreting non-verbal manifestations of depression.
• Evaluation of algorithms using datasets from the Audio/Visual Emotion Challenge (2013, 2014).
• Testing in gender-independent and gender-based modes with different fusion methods and classifiers.

Main Results:

• The proposed framework achieved a precision of 94.8% in detecting individuals with high scores on depressive symptomatology self-report scales.
• Optimal performance was achieved using a nearest neighbor classifier.
• Decision fusion of geometrical features (gender-independent) and audio features (gender-based) combined with an OR binary operation yielded the best results.

Conclusions:

• The developed system demonstrates high precision in detecting depression through non-verbal cues.
• Facial expression geometry and speech analysis are effective modalities for depression detection.
• The proposed framework shows promise as a decision support system for mental health professionals.