Emotional Body Movements for Automatic Depression Detection

This dataset contains three-dimensional skeletal motion data of emotional body movements (EBMs) collected for the study “The Impact of Depression on Emotional Body Motion for Automatic Depression Detection.” The dataset was designed to examine how depressive tendency influences bodily emotional expression and to support research on automatic depression detection (ADD) using body-motion signals.

A total of 63 participants were included and categorized into two groups based on depressive tendency: a Depression / Depressive Tendency group (n = 28) and a Control group (n = 35). Each participant performed posed bodily expressions of seven emotional states: happiness, sadness, surprise, anger, disgust, fear, and neutral. For each emotion, participants completed 20 repetitions, resulting in a total of 8,820 emotional body-motion samples (63 participants × 7 emotions × 20 trials).

The bodily movements were recorded using an RGB video camera and reconstructed into 3D skeletal sequences using a monocular 3D pose-estimation pipeline. Each motion sample consists of a 5-second sequence with 150 frames, represented by 17 body joints and 3 coordinate axes. The skeleton data were normalized by setting the initial hip position to the origin to ensure spatial consistency across participants.

The dataset is organized into two main folders:

Depression/: motion data from participants in the depressive tendency group

Control/: motion data from participants in the control group

Each CSV file corresponds to one emotional body-motion trial. The file names follow the format:

{Participant Number}_{Block}_{Trial}_{Emotion}.csv

where the emotion label is encoded as:

1 = Happy2 = Sad3 = Surprised4 = Angry5 = Disgust6 = Fearful7 = Neutral

For example:

01_1_3_1.csv

indicates a motion sample from Participant 01, Block 1, Trial 3, expressing Happiness.

Each CSV file contains the 3D skeletal coordinates of the emotional body motion sequence. Rows correspond to time frames, and columns correspond to the x-, y-, and z-coordinates of the 17 body joints. These data can be used for bodily emotion recognition, bodily motion map analysis, depression-related movement analysis, cross-emotional modeling, and automatic depression detection.

This dataset extends prior work on Bodily Motion Maps (BMMs) by providing emotional body-motion data from both control participants and participants with depressive tendency. It enables researchers to investigate depression-related changes in bodily emotional expressivity, inter-emotional movement contrast, and the feasibility of body-motion-based mental health assessment.

Usage Notes

Please note that the folder label “Depression” refers to the depressive tendency group defined by questionnaire-based screening and BDI criteria in the associated study. It should not be interpreted as a clinical diagnosis of major depressive disorder unless otherwise specified in the associated manuscript.

Citation

If you use this dataset, please cite the associated paper: Cho, Y., Jung, M., & Kim, K. (in preparation). The Impact of Depression on Emotional Body Motion for Automatic Depression Detection.