Depression Classification Using n-Gram Speech Errors from Manual and Automatic Stroop Color Test Transcripts.

Area of Science:

• Cognitive psychology
• Computational psychiatry
• Speech analysis

Background:

• The Stroop color test is widely used for temporal cognitive processing analysis.
• Limited research exists on verbal response pattern differences in depression.
• Sequential speech error patterns for depression classification remain unexplored.

Purpose of the Study:

• To analyze Stroop color test error patterns in clinically depressed versus healthy populations.
• To develop and evaluate speech error features for automatic depression classification.
• To investigate the efficacy of n-gram error sequences for depression detection.

Main Methods:

• Speech data collected via smart devices.
• Analysis of n-gram error sequence distributions from Stroop test responses.
• Utilized manual and automatic speech recognition (ASR) transcripts.
• Compared n-gram error features against an acoustic feature baseline.

Main Results:

• Depressed individuals exhibit significantly more Stroop color test errors, especially sequential ones.
• Trigram error features from manual transcripts achieved up to 95% depression classification accuracy.
• N-gram error features from ASR transcripts reached up to 90% accuracy.
• Acoustic features alone achieved only above 75% accuracy.

Conclusions:

• Sequential Stroop test error patterns are a promising digital biomarker for depression.
• N-gram error features derived from speech show high accuracy in depression classification.
• Smart device-based speech analysis offers a scalable approach for mental health screening.