A gender-emotion interaction multi-task network for depression recognition via transformer-based multimodal fusion

1

Depression significantly impacts individuals' work and life, characterized by high prevalence, recurrence, disability, and mortality.
2

Speech-based features are increasingly utilized for depression detection due to their non-invasive nature and ability to convey emotional states.
3

The proposed G-EIMTNet integrates gender and emotion in a multi-task framework to enhance depression recognition through transformer-based fusion.
4

Experiments demonstrated that G-EIMTNet outperformed baseline models by 15.88% in accuracy and 14.73% in F1 score on the AVEC2014 dataset.
5

The study highlights the importance of addressing gender and emotion in depression recognition to improve model robustness and accuracy.

Frontiers In Psychiatry

1