BASENet: Band-Adapted Speech Enhancement Network with Cross-Band Attention
Researchers have developed BASENet, a novel speech enhancement network that adapts its processing capacity based on the perceptual importance of different frequency bands. This architecture assigns more resources to lower frequencies, which are more critical for human hearing, and fewer to higher frequencies. The network incorporates a cross-band attention mechanism to capture harmonic relationships between frequency bands efficiently. BASENet demonstrates state-of-the-art performance with significantly fewer parameters and computational resources compared to existing methods, making it suitable for real-time applications on devices with limited capabilities. AI
IMPACT This novel architecture could lead to more efficient and effective real-time speech enhancement systems, particularly for resource-constrained devices.