WQ-Fusion: Dynamic Gated Attention for Cross-Domain Audio Representation

Mingda Lin, Lei Ding, Xinyue Zhou +6

While pre-trained models excel in specialized tasks, learning universal representations across diverse acoustic domains remains challenging. To address this, we propose WQ-Fusion,…