小数据下情绪识别，领域知识比复杂模型更有效
Attention Isn't All You Need for Emotion Recognition:Domain Features Outperform Transformers on the EAV Dataset
- 用领域适配的特征工程替代复杂注意力机制
- 音频加差分梅尔特征使准确率提升3.66个百分点
- 适合小样本情绪分析研究者参考
我们对EAV数据集上的多模态情绪识别进行了系统研究，检验复杂注意力机制在小数据集上的有效性。实验包含三类模型：基础Transformer（M1）、新型分解注意力机制（M2）和改进的CNN基线（M3）。结果表明，复杂注意力机制在小数据上持续表现不佳：M2模型准确率比基线低5至13个百分点，因过拟合和破坏预训练特征所致。相比之下，简单且符合领域的改进方法更有效：在音频CNN中加入差分梅尔倒谱系数（delta MFCCs），准确率从61.9%提升至65.56%（+3.66pp）；EEG采用频域特征后达到67.62%（较论文基线+7.62pp）。视觉Transformer基线（M1）通过领域特定预训练达75.30%，超过论文中ViViT的74.5%；加入视觉差分特征后达72.68%（较论文CNN基线+1.28pp）。这些发现表明，在小规模情绪识别任务中，领域知识与合理实现优于架构复杂性。
原文摘要 · Abstract (English)
We present a systematic study of multimodal emotion recognition using the EAV dataset, investigating whether complex attention mechanisms improve performance on small datasets. We implement three model categories: baseline transformers (M1), novel factorized attention mechanisms (M2), and improved CNN baselines (M3). Our experiments show that sophisticated attention mechanisms consistently underperform on small datasets. M2 models achieved 5 to 13 percentage points below baselines due to overfitting and destruction of pretrained features. In contrast, simple domain-appropriate modifications proved effective: adding delta MFCCs to the audio CNN improved accuracy from 61.9% to 65.56% (+3.66pp), while frequency-domain features for EEG achieved 67.62% (+7.62pp over the paper baseline). Our vision transformer baseline (M1) reached 75.30%, exceeding the paper's ViViT result (74.5%) through domain-specific pretraining, and vision delta features achieved 72.68% (+1.28pp over the paper CNN). These findings demonstrate that for small-scale emotion recognition, domain knowledge and proper implementation outperform architectural complexity.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。