1,835 gait sequences categorized into four emotion classes including Happy, Sad, Angry, and Neutral for affective computing research. The data consists of spatial-temporal graphs representing 16 skeletal joints across multiple frames of human walking to capture movement dynamics.
Use Cases
- Train a Spatial Temporal Graph Convolutional Network (ST-GCN) to predict emotion labels from skeletal joint sequences.
- Perform kinematic analysis of gait patterns by comparing joint trajectories across the four emotion classes.
- Develop feature extraction algorithms that identify the most significant skeletal joints for emotion perception using the graph node data.
Strengths
- 1,835 annotated gait sequences across 4 distinct emotion categories.
- 16 skeletal joint coordinates per frame represented as graph nodes.
- Spatial-temporal graph format capturing both joint connectivity and temporal movement sequences.
- Labels include Happy, Sad, Angry, and Neutral affective states.