Estimating Depressive Symptom Class from Voice.

Area of Science:

• Psychiatry
• Computational Linguistics
• Speech Analysis

Background:

• Voice-based methods offer an objective approach to depression detection.
• Current research primarily focuses on estimating depression presence or severity.
• Estimating specific depression symptoms is crucial for effective treatment and patient distress relief.

Purpose of the Study:

• To cluster depression symptoms using patient speech.
• To estimate patient symptom groups based on acoustic speech features.
• To assess the feasibility of voice analysis for detailed depression symptom profiling.

Main Methods:

• Collected speech data from patients diagnosed with depression.
• Utilized acoustic features extracted from patient speech recordings.
• Applied clustering algorithms to group patients based on speech characteristics.
• Correlated speech-based clusters with Hamilton Depression Rating Scale (HAM-D) scores.

Main Results:

• Successfully clustered patients into different symptom groups using speech.
• Achieved a classification accuracy of 79% in separating symptom groups.
• Demonstrated a significant correlation between acoustic features and depression symptom clusters.

Conclusions:

• Voice analysis is a viable tool for estimating depression symptom clusters.
• Speech-based symptom profiling can complement traditional depression assessment.
• This method holds potential for personalized depression treatment strategies.