Music Source Separation Based on a Lightweight Deep Learning Framework (DTTNET: DUAL-PATH TFC-TDF UNET)

TL;DR

This paper introduces DTTNet, a lightweight deep learning model for music source separation that achieves competitive results with significantly fewer parameters, emphasizing efficiency and generalization.

Contribution

The paper presents a novel lightweight architecture, DTTNet, combining dual-path modules and TFC-TDF UNet, improving efficiency while maintaining high separation quality.

Findings

DTTNet achieves 10.12 dB cSDR on vocals.

DTTNet uses 86.7% fewer parameters than BSRNN.

DTTNet demonstrates strong pattern-specific performance and generalization.

Abstract

Music source separation (MSS) aims to extract 'vocals', 'drums', 'bass' and 'other' tracks from a piece of mixed music. While deep learning methods have shown impressive results, there is a trend toward larger models. In our paper, we introduce a novel and lightweight architecture called DTTNet, which is based on Dual-Path Module and Time-Frequency Convolutions Time-Distributed Fully-connected UNet (TFC-TDF UNet). DTTNet achieves 10.12 dB cSDR on 'vocals' compared to 10.01 dB reported for Bandsplit RNN (BSRNN) but with 86.7% fewer parameters. We also assess pattern-specific performance and model generalization for intricate audio patterns.