Conditioned-U-Net: Introducing a Control Mechanism in the U-Net for Multiple Source Separations

Conditioned-U-Net: Introducing a Control Mechanism in the U-Net for Multiple Source Separations
复制标题

Conditioned-U-Net:在 U-Net 中引入控制机制以实现多源分离

DOI:
--
复制
发表时间:
2019
期刊:
International Society for Music Information Retrieval Conference
影响因子:
--
通讯作者:
Geoffroy Peeters
Geoffroy Peeters
中科院分区:
--
文献类型:
--
作者:
Gabriel Meseguer;Geoffroy Peeters

文献摘要

参考文献

被引文献

相似文献

用于音频源分离的数据驱动模型(例如U-Net或Wave-U-Net)通常是专用于单个任务并专门训练的模型,例如特定的乐器隔离。同时训练他们完成各种任务通常会导致比训练他们完成单一专业任务更糟糕的表现。在这项工作中,我们介绍了条件U网(C-U-Net),它增加了一个控制机制,以标准的U网。控制机制允许我们训练一个独特的通用U-Net来执行各种仪器的分离。C-U-Net根据独热编码输入向量决定要隔离的仪器。输入向量被嵌入以获得控制逐行线性调制(FilM)层的参数。Film层修改U-Net特征图,以便通过仿射变换分离所需的仪器。C-U-Net可以执行不同的仪器分离,所有这些都是通过一个单一的模型实现的,与专用的模型具有相同的性能,并且成本更低。
Data-driven models for audio source separation such as U-Net or Wave-U-Net are usually models dedicated to and specifically trained for a single task, e.g. a particular instrument isolation. Training them for various tasks at once commonly results in worse performances than training them for a single specialized task. In this work, we introduce the Conditioned-U-Net (C-U-Net) which adds a control mechanism to the standard U-Net. The control mechanism allows us to train a unique and generic U-Net to perform the separation of various instruments. The C-U-Net decides the instrument to isolate according to a one-hot-encoding input vector. The input vector is embedded to obtain the parameters that control Feature-wise Linear Modulation (FiLM) layers. FiLM layers modify the U-Net feature maps in order to separate the desired instrument via affine transformations. The C-U-Net performs different instrument separations, all with a single model achieving the same performances as the dedicated ones at a lower cost.
相位估计对基于时频掩蔽的单通道语音分离的影响。
DOI: 10.1121/1.4986647
发表时间: 2017
期刊: The Journal of the Acoustical Society of America
影响因子: --
作者:
Mayer,Florian;Williamson,DonaldS;Mowlaee,Pejman;Wang,DeLiang
通讯作者: Wang,DeLiang