Text-guided multimodal depression detection via cross-modal feature reconstruction and decomposition

Ziqiang Chen, Dandan Wang, Liangliang Lou, Shiqing Zhang, Xiaoming Zhao, Shuqiang Jiang, Jun Yu, Jun Xiao

Abstract: Highlights•Proposing a text-guided multimodal depression detection method.•Proposing a cross-modal feature reconstruction and decomposition framework.•Proposing a bi-directional cross attention module for feature enhancement.•Extensive experiments show the effectiveness of the proposed method.