使用 Hydra 改进 DF-Conformer 以实现基于离散编解码器 Token 的高保真生成式语音增强
Improving DF-Conformer Using Hydra For High-Fidelity Generative Speech Enhancement on Discrete Codec Token
AI总结：
本文提出用双向选择性结构化状态空间模型 Hydra 替换 FAVOR+，以改进 DF-Conformer 在语音增强中的全局序列建模并保持线性复杂度，在 Genhancer 模型上取得了更优性能。
摘要：
Dilated FAVOR Conformer (DF-Conformer) 是 Conformer 架构的高效变体，专为语音增强 (SE) 设计。它通过正交随机特征 (FAVOR+) 采用快速注意力机制，以缓解自注意力带来的二次复杂度，同时利用空洞卷积扩大感受野。这种组合在各种 SE 模型中表现出色。本文提出用双向选择性结构化状态空间序列模型替换 FAVOR+，以实现两个主要目标：(1) 消除 FAVOR+ 固有的近似以增强全局序列建模，(2) 保持相对于序列长度的线性复杂度。具体而言，我们使用 Hydra（Mamba 的双向扩展），置于结构化矩阵混合器框架内。在离散编解码器 token 上使用名为 Genhancer 的生成式 SE 模型进行的实验表明，所提方法超越了 DF-Conformer 的性能。
英文摘要：
The Dilated FAVOR Conformer (DF-Conformer) is an efficient variant of the Conformer architecture designed for speech enhancement (SE). It employs fast attention through positive orthogonal random features (FAVOR+) to mitigate the quadratic complexity associated with self-attention, while utilizing dilated convolution to expand the receptive field. This combination results in impressive performance across various SE models. In this paper, we propose replacing FAVOR+ with bidirectional selective structured state-space sequence models to achieve two main objectives:(1) enhancing global sequential modeling by eliminating the approximations inherent in FAVOR+, and (2) maintaining linear complexity relative to the sequence length. Specifically, we utilize Hydra, a bidirectional extension of Mamba, framed within the structured matrix mixer framework. Experiments conducted using a generative SE model on discrete codec tokens, known as Genhancer, demonstrate that the proposed method surpasses the performance of the DF-Conformer.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
