Lasaft: Latent Source Attentive Frequency Transformation For Conditioned Source Separation
Lasaft: Latent Source Attentive Frequency Transformation For Conditioned Source Separation
复制标题
Lasaft：用于条件源分离的潜在源关注频率变换
DOI：
10.1109/icassp39728.2021.9413896
复制
发表时间：
2020
期刊：
影响因子：
--
通讯作者：
Soonyoung Jung
中科院分区：
文献类型：
--
作者：
Woosung Choi;Minseok Kim;Jaehwa Chung;Soonyoung Jung
最近的深度学习方法表明，频率变换（FT）模块可以通过捕获频率模式来显着改进基于频谱图的单源分离模型。本文的目标是扩展 FT 模块以适应多源任务。我们提出潜在源关注频率变换（LaSAFT）块来捕获源相关频率模式。我们还提出了门控逐点卷积调制（GPoCM），它是特征线性调制（FiLM）的扩展，用于调制内部特征。通过采用这两种新颖的方法，我们扩展了 Conditioned-U-Net (CUNet) 进行多源分离，实验结果表明我们的 LaSAFT 和 GPoCM 可以提高 CUNet 的性能，在多个 MUSDB18 源分离任务上实现最先进的 SDR 性能。
Recent deep-learning approaches have shown that Frequency Transformation (FT) blocks can significantly improve spectrogram-based single-source separation models by capturing frequency patterns. The goal of this paper is to extend the FT block to fit the multi-source task. We propose the Latent Source Attentive Frequency Transformation (LaSAFT) block to capture source-dependent frequency patterns. We also propose the Gated Point-wise Convolutional Modulation (GPoCM), an extension of Feature-wise Linear Modulation (FiLM), to modulate internal features. By employing these two novel methods, we extend the Conditioned-U-Net (CUNet) for multi-source separation, and the experimental results indicate that our LaSAFT and GPoCM can improve the CUNet’s performance, achieving state-of-the-art SDR performance on several MUSDB18 source separation tasks.