Identifying Depressed from healthy cases using speech processing

As the emotion can affect the speech signal, we can extract a lot information with processing this signal. In this study we use speech signal to analysis the prosodic, vocal effects and glottal features for distinguish depress and healthy students. A new database of students with and without depressive disorder and treated depress students has collected. We extract the prosodic features (pitch and energy), vocal effect (formants) and glottal features. In present study, support vector machine (SVM) is used to classify the data. Two kinds of texts, emotional and scientific, are used to be read by human cases. Results indicate that scientific text speech is working better than emotional speech. In addition, our experiments show that proposed treatment protocol which was done by an expert psychologist has been effective to improve depression toward health.

Keywords:

depression , speech analysis , glottal , prosodics , vocal tract