Improving singing voice separation with the Wave-U-Net using Minimum Hyperspherical Energy
This work addresses the problem of extracting vocals from mixed music recordings for audio processing applications, representing an incremental improvement over existing methods.
The authors tackled singing voice separation by applying Minimum Hyperspherical Energy regularization to the Wave-U-Net, resulting in consistent improvements in Signal to Distortion Ratio and achieving the best time-domain system for this task.
In recent years, deep learning has surpassed traditional approaches to the problem of singing voice separation. The Wave-U-Net is a recent deep network architecture that operates directly on the time domain. The standard Wave-U-Net is trained with data augmentation and early stopping to prevent overfitting. Minimum hyperspherical energy (MHE) regularization has recently proven to increase generalization in image classification problems by encouraging a diversified filter configuration. In this work, we apply MHE regularization to the 1D filters of the Wave-U-Net. We evaluated this approach for separating the vocal part from mixed music audio recordings on the MUSDB18 dataset. We found that adding MHE regularization to the loss function consistently improves singing voice separation, as measured in the Signal to Distortion Ratio on test recordings, leading to the current best time-domain system for singing voice extraction.