Perceptual details of depression detection in Romanian language

Automatic detection of a depression state, studying voice inflexions and speech prosody, is carried out in this paper, using data extracted from the vocalic analysis of formants. We've done an important number of tests for depression detection using a K-Nearest Neighborhood algorithm run upon two databases with professional and normal voices recorded on selected texts in Romanian language. The selected stories were relevant for generating negative emotion (in our case depression). Our tests gave interesting results which have a large variety of possible applications.