Abstract: Effective detection of depression is crucial for early identification of potential risks. Traditional methods rely on clinical interviews, which are highly subjective and prone to misdiagnosis and delayed treatment. With the development of artificial intelligence, Deep Learning (DL) technology, with its superior feature representation and modeling capabilities, is gradually becoming a core method for depression detection. However, most existing reviews still focus on single-modality or traditional methods, lacking a systematic summary of multimodal deep learning methods in the field of depression detection, while multimodal data has significant advantages in providing complementary depressive cues. Therefore, this paper systematically reviews typical deep learning models and multimodal fusion strategies, summarizing the core technologies and latest advancements in this field. First, it reviews the evolution of depression detection methods from single-modality to multimodal, forming a development trajectory based on datasets and methods. Second, it focuses on discussing popular deep learning models and multimodal fusion strategies in depression detection applications, comparing the performance differences of various models and fusion strategies, and reviewing the advantages and limitations of different multimodal fusion techniques. Finally, it summarizes the challenges facing this field and potential directions for future innovation.

Key words: depression detection, multimodal data, fusion strategy, deep learning, feature fusion

摘要： 有效的抑郁症检测对于早期发现潜在风险至关重要。传统方法依赖于临床访谈具有较强的主观性，易导致误诊和延误治疗。随着人工智能的发展，深度学习（DL）技术凭借卓越的特征表示与建模能力，逐渐成为抑郁检测的核心方法。然而，现有综述大多仍侧重于单一模态或传统方法，缺乏对抑郁检测领域中多模态深度学习方法的系统归纳，而多模态数据在提供互补抑郁线索上具有显著优势。因此，系统地综述了典型深度学习模型与多模态融合策略，总结了该领域的核心技术和最新进展。首先，回顾了抑郁检测方法从单模态到多模态的演进过程，形成以数据集与方法为主线的发展脉络。其次，重点讨论了抑郁检测应用中流行的深度学习模型及多模态融合策略，对比了不同模型与融合策略的性能差异，同时评述了不同多模态融合技术的优势和局限性。最后，总结了该领域所面临的挑战与未来创新的潜在方向。

关键词: 抑郁检测, 多模态数据, 融合策略, 深度学习, 特征融合