Diagnostic Criteria for Depression based on Both Static and Dynamic Visual Features

Abstract

The mood disease depression is quite severe. Those who suffer from depression are often unable to function normally and may even resort to suicide if their condition worsens. Clinical interviews and questionnaires are now used in all cases of depression diagnosis, although these procedures are very subjective and lack objectivity and physiological basis. By calculating Beck Depression Inventory II (BDI-II) values from video data, we present an objective and non-discriminatory technique for depression diagnosis in this study. First, we use the LBP-TOP and EVLBP algorithms to extract a dynamic feature from each frame of the movie separately. The LBP operator is applied to each frame, HOG features are extracted from the LBP picture, and finally the LBP-HOG features are transformed into histogram vectors using BOW. Finally, the Gradient Boosting Regression is used to the combined dynamic and static characteristics to calculate the BDI-II. Using the AVEC 2014 depression dataset as an example, our tests demonstrate the efficacy of our suggested method.