CWS-PResUNet: Music Source Separation with Channel-wise Subband Phase-aware ResUNet

9 Dec 2021  ·  Haohe Liu, Qiuqiang Kong, Jiafeng Liu ·

Music source separation (MSS) shows active progress with deep learning models in recent years. Many MSS models perform separations on spectrograms by estimating bounded ratio masks and reusing the phases of the mixture. When using convolutional neural networks (CNN), weights are usually shared within a spectrogram during convolution regardless of the different patterns between frequency bands. In this study, we propose a new MSS model, channel-wise subband phase-aware ResUNet (CWS-PResUNet), to decompose signals into subbands and estimate an unbound complex ideal ratio mask (cIRM) for each source. CWS-PResUNet utilizes a channel-wise subband (CWS) feature to limit unnecessary global weights sharing on the spectrogram and reduce computational resource consumptions. The saved computational cost and memory can in turn allow for a larger architecture. On the MUSDB18HQ test set, we propose a 276-layer CWS-PResUNet and achieve state-of-the-art (SoTA) performance on vocals with an 8.92 signal-to-distortion ratio (SDR) score. By combining CWS-PResUNet and Demucs, our ByteMSS system ranks the 2nd on vocals score and 5th on average score in the 2021 ISMIR Music Demixing (MDX) Challenge limited training data track (leaderboard A). Our code and pre-trained models are publicly available at: https://github.com/haoheliu/2021-ISMIR-MSS-Challenge-CWS-PResUNet

PDF Abstract

Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Benchmark
Music Source Separation MUSDB18 CWS-PResUNet SDR (vocals) 8.92 # 7
SDR (drums) 6.38 # 21
SDR (other) 5.84 # 7
SDR (bass) 5.93 # 15
SDR (avg) 6.77 # 11
Music Source Separation MUSDB18-HQ CWS-PResUNet SDR (drums) 6.38 # 11
SDR (bass) 5.93 # 11
SDR (others) 5.84 # 10
SDR (vocals) 8.92 # 10
SDR (avg) 6.77 # 11

Methods