Paper Abstract and Keywords |
Presentation |
Improving Accuracy of Voice-Based Mental Health Evaluation by Bayesian Inference. Takeshi Takano, Yasuhiro Omiya, Uraguchi Tomotaka (PST), Masakazu Higuchi, Mitsuteru Nakamura, Shuji Shinohara, Shunji Mitsuyoshi (UTokyo), Taku Saito, Aihide Yoshino, Hiroyuki Toda (NDMC), Shinichi Tokuno (UTokyo) |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
Early detection of mental health malfunction is an issue in today's stressful modern era. We have developed MIMOSYS (Mind Monitoring System) which estimates human's mental health condition by voice. This technology has the advantage of being noninvasive and easy to measure. On the other hand, MIMOSYS has a property that the specificity is low at estimating mental health status from short-term speech. In this study, we collected speech data of patients with major depressive disorder, found the likelihood distribution of Vitality of mild and severe individuals, calculated the posterior probability using Bayes' theorem from Vitality which obtained from subject's utterance and estimated patients’ severity. As a result, it was confirmed that the separation performance improves more than Vitality itself. From the result, it was shown that it is possible to estimate mental health condition with high precision from short-term MIMOSYS data. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
Voice / Pathophysiology / Distinguish / Major Depressive Disorder / / / / |
Reference Info. |
IEICE Tech. Rep. |
Paper # |
|
Date of Issue |
|
ISSN |
|
Download PDF |
|