2-04 - Investigating U-nets with Various Intermediate Blocks for Spectrogram-based Singing Voice Separation

Woosung Choi, Minseok Kim, Jaehwa Chung, Daewon Lee, Soonyoung Jung

Keywords: MIR tasks, Sound source separation, Musical features and properties, Timbre, instrumentation, and voice

Abstract: Singing Voice Separation (SVS) tries to separate singing voice from a given mixed musical signal. Recently, many U-Net-based models have been proposed for the SVS task, but there were no existing works that evaluate and compare various types of intermediate blocks that can be used in the U-Net architecture. In this paper, we introduce a variety of intermediate spectrogram transformation blocks. We implement U-nets based on these blocks and train them on complex-valued spectrograms to consider both magnitude and phase. These networks are then compared on the SDR metric. When using a particular block composed of convolutional and fully-connected layers, it achieves state-of-the-art SDR on the MUSDB singing voice separation task by a large margin of 0.9 dB. Our code and models are available online.