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

2025-06-17

AI Summary

Addressing the challenge of balancing noise suppression, perceptual quality, and speaker identity preservation in real-world noisy environments, this work presents the first unified benchmark evaluation of Wave-U-Net, CMGAN, and U-Net on three authentic multi-source noise datasets: SpEAR, VPQAD, and Clarkson. We employ a comprehensive set of objective metrics—SNR, PESQ, and VeriSpeak—to systematically assess model performance across these dimensions. Results reveal complementary strengths: U-Net achieves superior noise suppression (364.2% SNR improvement on Clarkson), CMGAN attains the highest perceptual quality (PESQ = 4.04 on SpEAR), and Wave-U-Net excels in speaker identity preservation (27.38% VeriSpeak improvement on VPQAD). This study establishes a reproducible, empirically grounded benchmark that informs model selection and synergistic design for speech enhancement under realistic acoustic conditions.

Technology Category

• Machine Learning: Multimodal Learning
• Natural Language Processing: Speech
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Web Mining and Content Analysis: Large pretrained models with web data
• User Modeling, Personalization and Recommendation: Metrics for user behavior and evaluating success

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.

Problem

• Evaluates deep learning models for speech enhancement in noisy environments
• Compares noise suppression, perceptual quality, and speaker feature retention
• Benchmarks Wave-U-Net, CMGAN, and U-Net on diverse real-world datasets

Innovation

• Benchmarks Wave-U-Net, CMGAN, and U-Net models
• U-Net excels in noise suppression performance
• CMGAN achieves highest perceptual quality scores