Moises-Light: Resource-efficient Band-split U-Net For Music Source Separation

2025-10-08

AI Summary

To address the challenge of deploying computationally expensive music source separation models on resource-constrained devices, this paper proposes a lightweight band-split U-Net architecture. The method introduces band-split convolution and a dual-path feature fusion mechanism within the U-Net encoder-decoder framework, enabling frequency-band-adaptive modeling and efficient cross-path information interaction. Despite its significantly reduced parameter count—13× fewer than state-of-the-art large models—the proposed architecture achieves competitive source separation performance, matching top-performing models in Signal-to-Distortion Ratio (SDR) on MUSDB-HQ. Moreover, it demonstrates strong generalization and scalability on the extended MoisesDB dataset. This work substantially raises the performance ceiling for lightweight source separation models and provides a viable solution for real-time, on-device music separation.

Technology Category

• Machine Learning: Mixture of Experts (MoE)
• Computer Vision: Diffusion Models for Vision
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

In recent years, significant advances have been made in music source separation, with model architectures such as dual-path modeling, band-split modules, or transformer layers achieving comparably good results. However, these models often contain a significant number of parameters, posing challenges to devices with limited computational resources in terms of training and practical application. While some lightweight models have been introduced, they generally perform worse compared to their larger counterparts. In this paper, we take inspiration from these recent advances to improve a lightweight model. We demonstrate that with careful design, a lightweight model can achieve comparable SDRs to models with up to 13 times more parameters. Our proposed model, Moises-Light, achieves competitive results in separating four musical stems on the MUSDB-HQ benchmark dataset. The proposed model also demonstrates competitive scalability when using MoisesDB as additional training data.

Problem

• Lightweight model for music source separation
• Reducing computational resource requirements
• Achieving competitive performance with fewer parameters

Innovation

• Resource-efficient band-split U-Net architecture
• Lightweight model with competitive separation performance
• Scalable design using additional training data