恢复、分离、恢复：音乐源恢复的模块化框架
Restore, Separate, Restore: A Modular Framework for Music Source Restoration
浏览论文内容
中文总结 AI 辅助
提出一个三阶段模块化框架，通过先恢复混合、再分离八个分轨、最后针对残余伪影微调恢复专家，解决音乐源恢复中反转非线性制作效果和传输退化的问题，并在 MSR 挑战测试集上逐步提升恢复质量。
中文摘要 AI 辅助
音乐源恢复（MSR）旨在从混合、母带处理且可能退化的录音中恢复原始未处理的乐器分轨。与传统的源分离（将混合视为干净源的线性叠加）不同，MSR 还必须反转非线性制作效果（如均衡和压缩）以及传输相关的退化（如编解码器伪影）。我们提出了一个围绕这一区别构建的三阶段框架：（1）一个混合恢复模型，在分离之前处理退化；（2）一个单一模型，将恢复后的混合分离成八个目标分轨（人声、吉他、键盘、合成器、贝斯、鼓、打击乐和管弦乐）；（3）针对分离器自身残余伪影进行微调的分轨特定恢复专家。在 MSR 挑战测试集上，每个阶段都比前一阶段提高了恢复质量。我们发布了代码和模型，以支持未来在 MSR 领域的研究，网址为 https://this URL。
英文摘要
Music Source Restoration (MSR) seeks to recover original, unprocessed instrument stems from mixed, mastered, and possibly degraded recordings. Unlike conventional source separation, which treats the mixture as a linear sum of clean sources, MSR must additionally invert nonlinear production effects, such as equalization and compression, and transmission-related degradations, such as codec artifacts. We propose a three-stage framework built around this distinction: (1) a mixture restoration model that addresses degradation before separation, (2) a single model that separates the restored mixture into eight target stems (vocals, guitars, keyboards, synthesizers, bass, drums, percussion, and orchestra), and (3) stem-specific restoration experts fine-tuned on the separator's own residual artifacts. Each stage improves restoration quality over the previous one on the MSR Challenge test set. We release code and models to support future research in MSR at https://github.com/theMoro/music_source_restoration.
发表机构
- Institute of Computational Perception, Johannes Kepler University Linz（林茨约翰·开普勒大学计算感知研究所）
- LIT Artificial Intelligence Lab（LIT人工智能实验室）
机构由 AI 辅助整理，请以论文原文为准。
