In multimodal emotion recognition, disentangled representation learning method effectively address the inherent heterogeneity among modalities. To facilitate the flexible integration of enhanced disentangled features into multimodal emotional features, we propose a task-driven multimodal emotion recognition method TDMER. Its Cross-Modal Learning module promotes adaptive cross-modal learning of features disentangled into modality-invariant and modality-specific subspaces, based on their contributions to emotional classification probabilities. The Task-Contribution Fusion mechanism then assigns controllable weights to the enhanced features according to their task objectives, generating multimodal fusion features that improve the emotion classifier's discriminative ability. The proposed TDMER approach has been evaluated on two widely-used multimodal emotion recognition benchmarks and demonstrated significant performance improvements compared with other state-of the-art methods.

Speech signals contain rich information, such as textual content, emotion, and speaker identity. To extract these features more efficiently, researchers are investigating joint training across multiple tasks, like Speech Emotion Recognition (SER) and Speaker Verification (SV), aiming to improve performance by decoupling task-specific knowledge. Traditional multitask decoupling methods in SER typically use orthogonalization to increase the distance between parameter vectors in the feature space. In this paper, we introduce a novel Hybrid instance-level Contrastive Decoupling Loss. This method leverages supervised labels to effectively decouple SER and SV. Unlike previous approaches, it is not restricted to dual-stream models with identical architectures and can be easily integrated with leading models for each sub-task. Experimental results show that our proposed Hybrid Contrastive Learning Decoupling (HCLD) method significantly outperforms traditional orthogonal decoupling approaches.