A Comparative Evaluation of Deep Learning Models for Speech Enhancement in Real-World Noisy Environments

TL;DR

This study benchmarks three deep learning models—Wave-U-Net, CMGAN, and U-Net—for speech enhancement in real-world noisy environments using standardized training and evaluation across DEMAND-based training data and SpEAR, VPQAD, and Clarkson evaluation sets. It integrates traditional quality metrics (PESQ, SNR) with biometric relevance via VeriSpeak to assess speaker-preservation alongside denoising. The results show U-Net delivers strongest noise suppression, CMGAN achieves the best perceptual quality, and Wave-U-Net offers a balanced profile with notable speaker-recognition gains, highlighting clear trade-offs among denoising, quality, and biometric fidelity. The findings inform model selection for forensic, telecommunication, and speaker-verification applications and underscore the value of real-world, multi-metric evaluation in guiding practical deployment.

Abstract

Speech enhancement, particularly denoising, is vital in improving the intelligibility and quality of speech signals for real-world applications, especially in noisy environments. While prior research has introduced various deep learning models for this purpose, many struggle to balance noise suppression, perceptual quality, and speaker-specific feature preservation, leaving a critical research gap in their comparative performance evaluation. This study benchmarks three state-of-the-art models Wave-U-Net, CMGAN, and U-Net, on diverse datasets such as SpEAR, VPQAD, and Clarkson datasets. These models were chosen due to their relevance in the literature and code accessibility. The evaluation reveals that U-Net achieves high noise suppression with SNR improvements of +71.96% on SpEAR, +64.83% on VPQAD, and +364.2% on the Clarkson dataset. CMGAN outperforms in perceptual quality, attaining the highest PESQ scores of 4.04 on SpEAR and 1.46 on VPQAD, making it well-suited for applications prioritizing natural and intelligible speech. Wave-U-Net balances these attributes with improvements in speaker-specific feature retention, evidenced by VeriSpeak score gains of +10.84% on SpEAR and +27.38% on VPQAD. This research indicates how advanced methods can optimize trade-offs between noise suppression, perceptual quality, and speaker recognition. The findings may contribute to advancing voice biometrics, forensic audio analysis, telecommunication, and speaker verification in challenging acoustic conditions.

Figures (4)

• Figure 1: Data Collection Setup for the Clarkson Dataset, Showing Participant with Dual Microphone Setup
• Figure 2: The Wave-U-Net Architecture Stoller2018Macartney2018
• Figure 3: Overview of the CMGAN Architecture Abdulatif2024_Arxiv
• Figure 4: Overview of the U-Net Architecture UNet2015