Fifth International Workshop on Models and Analysis of Vocal Emissions for Biomedical Applications (MAVEBA 2007)

Florence, Italy
December 13-15, 2007

Distinguishing High Risk Suicidal Subjects Among Depressed Subjects Using Mel-Frequency Cepstrum Coefficients and Cross Validation Technique

H. Kaymaz Keskinpala (1), T. Yingthawornsuk (1), D. Mitchell Wilkes (1), Richard G. Shiavi (1,2), R. M. Salomon (3)

(1) Department of Electrical and Computer Engineering; (2) Department of Biomedical Engineering; (3) Department of Psychiatry, School of Medicine
Vanderbilt University, Nashville, TN, USA

This paper describes a way to distinguish high risk suicidal patients among depressed patients using mel-frequency cepstrum coefficients. Distinguishing high risk suicidal patients among depressed patients is an important problem; a practical solution to this would prevent the loss of many lives. In this study, the vocal characteristics of female and male patients' speech samples were analyzed and a small subset of the first ten melfrequency cepstrum coefficients were used to classify high risk suicidal patients and depressed patients. Cross validation was used to observe classification performance. There were two different types of speech samples from both male and female patients. One of them was the speech sampled during a clinical interview and the other was speech sampled during a text-reading session.
Index Terms. Speech, MFCC, suicide, depression, cross validation

Full Paper (reprinted with permission from Firenze University Press)

Bibliographic reference.  Kaymaz Keskinpala, H. / Yingthawornsuk, T. / Wilkes, D. Mitchell / Shiavi, Richard G. / Salomon, R. M. (2007): "Distinguishing high risk suicidal subjects among depressed subjects using mel-frequency cepstrum coefficients and cross validation technique", In MAVEBA-2007, 157-160.