Neural-net classification for spatio-temporal descriptor based depression analysis

Depression is a severe psychiatric disorder. Despite the high prevalence, current clinical practice depends almost exclusively on self-report and clinical opinion, risking a range of subjective biases. This paper focuses on depression analysis based on visual cues from facial expressions and upper body movements. The proposed diagnostic support system is based on computing spatio-temporal features from video sequences. Space Time Interest Points are computed for the videos for analysing the upper body movements and a temporal visual words dictionary is learned from them. Intra-facial muscle movement is captured by computing a LBP-TOP based codebook. Various neural-net classifiers are explored and compared with a SVM. The approach is evaluated on real-world clinical data from interactive interviews with depressed and healthy subjects.