Article: Siamese network contrastive learning model for the objective evaluation of singing sound quality Journal: International Journal of Information and Communication Technology (IJICT) 2026 Vol.27 No.32 pp.18 - 33 Abstract: This study extracts target singers' voices from mixed audio with background music and noise, addressing the subjectivity, instability and lack of objective standards in traditional evaluation. An innovative SNN-SpEx+ method is proposed, combining Siamese neural network (SNN) and contrastive learning-based SpEx+. Its parameter-sharing twin architecture unifies the feature space for reference and mixed speech, breaking the bottleneck of feature space dislocation in traditional dual-network structures. Contrastive learning is integrated into vocal extraction to build a 'separation is learning' joint optimisation framework, enhancing adaptability to unknown singers and short reference voices. Experiments on MUSDB18-HQ and NSynth-Singer show SNN-SpEx+ outperforms SpEx+ + by 0.85 dB in SI-SDRi and 0.17 in PESQ. For short references (<2 s), its SI-SDRi drops only 3.16 dB (3.4 dB lower than SpEx++), providing an automatic standardised evaluation tool for music education and singer selection with broad prospects. Inderscience Publishers - linking academia, business and industry through research

Title: Siamese network contrastive learning model for the objective evaluation of singing sound quality

Authors: Xiaochen Liang

Addresses: School of Performance, Sichuan University of Media and Communications, Chengdu, 611745, China

Abstract: This study extracts target singers' voices from mixed audio with background music and noise, addressing the subjectivity, instability and lack of objective standards in traditional evaluation. An innovative SNN-SpEx+ method is proposed, combining Siamese neural network (SNN) and contrastive learning-based SpEx+. Its parameter-sharing twin architecture unifies the feature space for reference and mixed speech, breaking the bottleneck of feature space dislocation in traditional dual-network structures. Contrastive learning is integrated into vocal extraction to build a 'separation is learning' joint optimisation framework, enhancing adaptability to unknown singers and short reference voices. Experiments on MUSDB18-HQ and NSynth-Singer show SNN-SpEx+ outperforms SpEx+ + by 0.85 dB in SI-SDRi and 0.17 in PESQ. For short references (<2 s), its SI-SDRi drops only 3.16 dB (3.4 dB lower than SpEx++), providing an automatic standardised evaluation tool for music education and singer selection with broad prospects.

Keywords: singing; sound quality; vocal extraction; Siamese neural network; SNN; contrastive learning.

DOI: 10.1504/IJICT.2026.152862

International Journal of Information and Communication Technology, 2026 Vol.27 No.32, pp.18 - 33

Received: 31 Oct 2025
Accepted: 12 Dec 2025
Published online: 13 Apr 2026 *

Title: Siamese network contrastive learning model for the objective evaluation of singing sound quality

Keep up-to-date