TS-URGENet: A Three-stage Universal Robust and Generalizable Speech Enhancement Network

2025/05/24 by Xiaobin Rong, Da‐Han Wang, Rong, Xiaobin +9 · 2 citations
Computer Science · Health Professions · #Artificial Intelligence (cs.AI) #Audio and Speech Processing (eess.AS) #Background noise #Clipping (morphology) #Codec #Distortion (music) #FOS: Computer and information sciences #FOS: Electrical engineering #Infant Health and Development #Packet loss #Residual #Speech Recognition and Synthesis #Speech and Audio Processing #Speech coding #Speech enhancement #Voice activity detection #electronic engineering #information engineering

paper · pdf · doi:10.48550/arxiv.2505.18533

published in arXiv (Cornell University) (Cornell University)

openalex publication_date 2025/05/24 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/05

Abstract

Universal speech enhancement aims to handle input speech with different distortions and input formats. To tackle this challenge, we present TS-URGENet, a Three-Stage Universal, Robust, and Generalizable speech Enhancement Network. To address various distortions, the proposed system employs a novel three-stage architecture consisting of a filling stage, a separation stage, and a restoration stage. The filling stage mitigates packet loss by preliminarily filling lost regions under noise interference, ensuring signal continuity. The separation stage suppresses noise, reverberation, and clipping distortion to improve speech clarity. Finally, the restoration stage compensates for bandwidth limitation, codec artifacts, and residual packet loss distortion, refining the overall speech quality. Our proposed TS-URGENet achieved outstanding performance in the Interspeech 2025 URGENT Challenge, ranking 2nd in Track 1.

Citations

• Interspeech 2025 URGENT Speech Enhancement Challenge
• KS-Net: Multi-band joint speech restoration and enhancement network for 2024 ICASSP SSI Challenge
• SpeechBERTScore: Reference-Aware Automatic Evaluation of Speech Generation Leveraging NLP Evaluation Metrics
• ICASSP 2024 Speech Signal Improvement Challenge
• Towards High-Quality and Efficient Speech Bandwidth Extension with Parallel Amplitude and Phase Prediction
• RaD-Net: A Repairing and Denoising Network for Speech Signal Improvement
• BS-PLCNet: Band-split Packet Loss Concealment Network with Multi-task Learning Framework and Multi-discriminators
• Toward Universal Speech Enhancement for Diverse Input Conditions
• ICASSP 2023 Speech Signal Improvement Challenge
• TF-GridNet: Integrating Full- and Sub-Band Modeling for Speech Separation
• BigVGAN: A Universal Neural Vocoder with Large-Scale Training
• Universal Speech Enhancement with Score-based Diffusion
• UTMOS: UTokyo-SaruLab System for VoiceMOS Challenge 2022
• WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processing
• DNSMOS P.835: A Non-Intrusive Perceptual Objective Speech Quality Metric to Evaluate Noise Suppressors
• Audio Codec Enhancement with Generative Adversarial Networks

Cited by

• TF-Restormer: Complex Spectral Prediction for Speech Restoration
• GAP-URGENet: A Generative-Predictive Fusion Framework for Universal Speech Enhancement

Related

• FUSE: Universal Speech Enhancement using Multi-Stage Fusion of Sparse Compression and Token Generation Models for the URGENT 2025 Challenge
• KS-Net: Multi-band joint speech restoration and enhancement network for 2024 ICASSP SSI Challenge
• FINALLY: fast and universal speech enhancement with studio-like quality
• EDNet: A Versatile Speech Enhancement Framework with Gating Mamba Mechanism and Phase Shift-Invariant Training
• From Continuous to Discrete: Cross-Domain Collaborative General Speech Enhancement via Hierarchical Language Models