Addressing Missing and Noisy Modalities in One Solution: Unified Modality-Quality Framework for Low-quality Multimodal Data
This work addresses the challenge of improving model robustness for multimodal affective computing in real-world scenarios where data quality is often poor, representing an incremental advance by unifying separate prior approaches.
The paper tackles the problem of low-quality multimodal data, including noisy and missing modalities, by proposing a unified modality-quality framework that jointly addresses both issues to enhance model robustness, achieving state-of-the-art performance across multiple datasets under various modality conditions.
Multimodal data encountered in real-world scenarios are typically of low quality, with noisy modalities and missing modalities being typical forms that severely hinder model performance and robustness. However, prior works often handle noisy and missing modalities separately. In contrast, we jointly address missing and noisy modalities to enhance model robustness in low-quality data scenarios. We regard both noisy and missing modalities as a unified low-quality modality problem, and propose a unified modality-quality (UMQ) framework to enhance low-quality representations for multimodal affective computing. Firstly, we train a quality estimator with explicit supervised signals via a rank-guided training strategy that compares the relative quality of different representations by adding a ranking constraint, avoiding training noise caused by inaccurate absolute quality labels. Then, a quality enhancer for each modality is constructed, which uses the sample-specific information provided by other modalities and the modality-specific information provided by the defined modality baseline representation to enhance the quality of unimodal representations. Finally, we propose a quality-aware mixture-of-experts module with particular routing mechanism to enable multiple modality-quality problems to be addressed more specifically. UMQ consistently outperforms state-of-the-art baselines on multiple datasets under the settings of complete, missing, and noisy modalities.