Facial geometry and speech analysis for depression detection

Abstract

Depression is one of the most prevalent mental disorders, burdening many people world-wide. A system with the potential of serving as a decision support system is proposed, based on novel features extracted from facial expression geometry and speech, by interpreting non-verbal manifestations of depression. The proposed system has been tested both in gender independent and gender based modes, and with different fusion methods. The algorithms were evaluated for several combinations of parameters and classification schemes, on the dataset provided by the Audio/Visual Emotion Challenge of 2013 and 2014. The proposed framework achieved a precision of 94.8% for detecting persons achieving high scores on a self-report scale of depressive symptomatology. Optimal system performance was obtained using a nearest neighbour classifier on the decision fusion of geometrical features in the gender independent mode, and audio based features in the gender based mode; single visual and audio decisions were combined with the OR binary operation.

Glottal Source Features for Automatic Speech-Based Depression Assessment

“…Potential developments of the proposed method include the inclusion of complementary features both within the audio domain and from the visual modality cf. [31,19].…”

“…The binary classification method of Statak et al [17] assigned BDI-II scores of [0-9] to 'none-low depression' and to 'moderate-high depression', and used Geneva Minimalistic Acoustic Parameter Set features [18] for estimation of arousal level, achieving a classification accuracy of 82.5%. Pampouchidou et al [19] proposed binary categorical assessment, by considering BDI-II scores [0-13] as not-depressed, and as depressed. Using the covarep toolbox [20] for extracting speech-based DCT features, they achieved an F1-score of 0.64 for gender-based depression classification.…”

“…Labels of the testing partition were withheld for challenge purposes, so in the present study only the 200 recordings from the development and train sets were used. The dataset was annotated with a single label per recording, corresponding to speakers' scores on BDI-II, which according to its standardized cutoffs can be interpreted as minimal depression for a score of [0-13], mild [14][15][16][17][18][19], moderate [20][21][22][23][24][25][26][27][28]…”

A computer vision based image processing system for depression detection among students for counseling

“…The results pointed out that the most accurate action unit for depression detection was AU14 (action unit related to contempt). In [2], the identification of depression was done by analysing facial landmark points. The distances between them were found out using euclidean and city block distance methods.…”

Quantitative comparison of motion history image variants for video-based depression assessment

“…Similar results were obtained in the crosscultural study of Alghowinem et al [15] who focused on geometrical features derived from eye activity to achieve 81.3% classification accuracy utilizing one subset of the AVEC'13 dataset, among other datasets. Pampouchidou et al [16] reported 74.5% accuracy with the Local Curvelet Binary Patterns in Pairwise Orthogonal Planes, while Pampouchidou et al [17] employed geometrical features to achieve an F1 score of 58.6% in the single-modality (i.e., visual) approach and 72.8% when taking into account both audio and visual features.…”