Recognizing Emotions Through Body Movement
Researchers from the Information Processing and Telecommunications Center (IPTC) from the Universidad Politécnica de Madrid (UPM) have developed a new approach to recognizing human emotions from body movements. The study introduces a Window Transformer model that analyses motion sequences by dividing them into overlapping temporal segments and identifying which parts contain the most relevant emotional information.
Unlike approaches that treat an entire movement sequence equally, the proposed method learns to focus on the moments that contribute most to recognizing an emotion. This also makes the system more interpretable, as it can indicate which parts of a movement sequence influenced its prediction.
The method uses information about body landmarks and how they change over time, including movement and acceleration. This allows the system to capture not only body posture but also the dynamics of movement. Experiments on two datasets showed improvements in emotion-recognition accuracy compared with a conventional Transformer model.
Because the model uses motion landmarks rather than facial appearance, audio, or scene information, it may be suitable for settings where these modalities are unavailable or undesirable. In camera-based implementations, however, privacy ultimately depends on how video is processed, stored, and protected.
The research provides benchmark results and an interpretable modelling approach that future work can build on for human–computer interaction, affective computing, assistive technologies, and other systems that use body movement to understand aspects of human behavior.
Bibliographic reference:
Esteban-Romero S, Martín-Fernández I, San-Segundo R, Gil-Martín M and Fernández-Martínez F (2026) Emotion recognition from body movement through interpretable motion-aware sequential modeling. Front. Artif. Intell. 9:1897208. https://doi.org/10.3389/frai.2026.1897208
Authors:
Sergio Estebán Romero: GS /ORCID/ LinkedIn
Iván Martín Fernández: GS /ORCID/ LinkedIn
Rubén San Segundo Hernández: GS /ORCID/ LinkedIn
Manuel Gil Martín: GS /ORCID/ LinkedIn
Fernando Fernández Martínez: GS /ORCID/ LinkedIn
LinkedIn: https://www.linkedin.com/company/iptc-upm/
For more information: www.iptc.upm.es
Share this:
