Classification of bipolar disorder, major depressive disorder, and healthy state using voice

Abstract

Objective: In this study, we propose a voice index to identify healthy individuals, patients with bipolar disorder, and patients with major depressive disorder using polytomous logistic regression analysis. Methods: Voice features were extracted from voices of healthy individuals and patients with mental disease. Polytomous logistic regression analysis was performed for some voice features. Results: With the prediction model obtained using the analysis, we identified subject groups and were able to classify subjects into three groups with 90.79% accuracy. Conclusion: These results show that the proposed index may be used as a new evaluation index to identify depression.