基于人耳听觉特性的语音增强网络，高效提升低频细节保留能力
BASENet: Band-Adapted Speech Enhancement Network with Cross-Band Attention
- 按听觉感知密度划分频带，低频用更深网络，高频用轻量结构
- 跨频带注意力机制在线性复杂度下捕捉谐波依赖关系
- 参数少、速度快，适合实时部署于资源受限设备
语音增强模型通常对所有频率使用相同容量，忽略了人耳听觉的非均匀频谱分辨率。我们提出BASENet，一种频带自适应架构：将频谱按巴克尺度分段，根据临界带密度为各频带分配缩放后的编码器容量，自动为感知密集的低频分配更深分支，高频则采用轻量结构。通过紧凑的线性池化表示，跨频带注意力模块以线性复杂度捕获频带间的谐波依赖。该模型基于反向残差块与密集连接，并结合卷积循环网络，在VoiceBank+DEMAND数据集上实现3.55 PESQ和STOI约96%，仅需0.83M参数和7.3 G MACs，是所有PESQ>3.50方法中参数最少者。因果版本（3.44 PESQ）超越多个非因果基线，证明其适用于资源受限设备的实时流式处理。
原文摘要 · Abstract (English)
Speech enhancement models typically apply uniform capacity across all frequencies, disregarding the non-uniform spectral resolution of human hearing. We propose BASENet, a frequency-adapted architecture that partitions the spectrum into Bark-scale bands and assigns each a scaled-capacity encoder derived from critical-band density, automatically granting deeper branches to perceptually dense low frequencies and lighter ones to high frequencies. A cross-band attention module captures harmonic dependencies across bands through compact frequency-pooled representations at linear complexity. Built on inverted residual blocks with dense connectivity and a convolutional recurrent network, BASENet achieves 3.55 PESQ and STOI~96% on VoiceBank+DEMAND with only 0.83M parameters and 7.3 G~MACs, the fewest parameters among all methods with PESQ > 3.50. A causal variant (3.44 PESQ) surpasses several non-causal baselines, confirming suitability for real-time streaming on resource-constrained devices.
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。