现代信息科技2025，Vol.9Issue(16)：57-62,69,7.DOI:10.19850/j.cnki.2096-4706.2025.16.011
时空变化注意力机制图神经网络的音频事件分类研究
Research on Audio Event Classification Based on Graph Neural Network with Spatio-Temporal Variation Attention Mechanism
张墨华 1刘霁1
作者信息
- 1. 河南财经政法大学 计算机与信息工程学院,河南 郑州 450046
- 折叠
摘要
Abstract
Audio event classification faces challenges in complex scenarios,and the existing methods struggle to capture temporal relationships effectively.To address this,this paper proposes a Spatio-Temporal Variation Attention based Graph Neural Network(STVA-GNN),which models audio-visual segments as sequential graph nodes and leverages a Negative Attention Mechanism to compute spatiotemporal variation features between adjacent nodes,enhancing intra-modal and cross-modal dynamic information interactions.The core innovations include that a Contextual Information Compensation Module(CICM)captures spatiotemporal evolution patterns,and a Cross-Modal Graph Variation Incentive Module(CMGVI)enhances audio node weights using video-modal spatiotemporal variations for deep fusion.Experimental results on the AudioSet dataset demonstrate that STVA-GNN achieves mAP and AUC scores of 0.56 and 0.94 respectively,outperforming mainstream methods.Additionally,it maintains a significant advantage in noisy environments,verifying its robustness.关键词
音频事件分类/时空变化注意力机制/时序图神经网络/变化信息补偿/跨模态信息融合Key words
audio event classification/Spatio-Temporal Variation Attention Mechanism/Temporal Graph Neural Network/change information compensation/cross-modal information fusion分类
信息技术与安全科学引用本文复制引用
张墨华,刘霁..时空变化注意力机制图神经网络的音频事件分类研究[J].现代信息科技,2025,9(16):57-62,69,7.
