论文解读
Towards Distance-Aware Synthetic Audio Mixtures for Universal Sound Separation
语音分离 | 6.5/10
语音分离
论文解读
Training Dynamics-Aware Multi-Factor Curriculum Learning for Target Speaker Extraction
语音分离 | 7.0/10
论文解读
UNMIXX: Untangling Highly Correlated Singing Voices Mixtures
语音分离 | 8.5/10
论文解读
Vib2Sound: Separation Of Multimodal Sound Sources
语音分离 | 6.5/10
论文解读
VM-UNSSOR: Unsupervised Neural Speech Separation Enhanced by Higher-SNR Virtual Microphone Arrays
语音分离 | 7.5/10
论文解读
Towards Streaming Target Speaker Extraction via Chunk-wise Interleaved Splicing of Autoregressive Language Model
1. **要解决什么问题**：现有基于生成模型（如扩散模型、自回归模型）的目标说话人提取（TSE）方法依赖全局上下文，难以直接用于实时流式场景，强行适配会导致性能严重下降。 2. **方法核心是什么**：提出首个面向流式TSE的自回归（AR）框架，核心是“分块交错拼接范式”。该范式将混合语音分块
论文解读
Towards Streaming Target Speaker Extraction via Chunk-wise Interleaved Splicing of Autoregressive Language Model
这篇论文旨在解决生成式目标说话人提取（TSE）模型在流式实时应用中因依赖全局上下文而导致性能严重下降的核心问题。作者首次提出了一个基于自回归语言模型（LauraGPT）的流式TSE框架。其核心创新是“分块交织拼接范式”，通过将混合音频块与对应的目标语音离散编码块交错排列作为模型输入，严格保证了推理的