MeanFlowSE: One-Step Generative Speech Enhancement via MeanFlow

2025-09-27

AI Summary

Existing generative speech enhancement methods—such as diffusion models and flow matching—rely on multi-step sampling or large-scale architectures, resulting in slow inference and poor deployability. To address this, we propose MeanFlowSE: the first method to introduce MeanFlow into speech enhancement, establishing a single-step latent-space optimization framework that generates enhanced speech in one pass by predicting the mean velocity field. Crucially, it replaces conventional VAE latent variables with self-supervised learning (SSL)-derived representations, providing robust semantic guidance. Evaluated on the DNS Challenge blind test set, MeanFlowSE achieves state-of-the-art perceptual quality (P.808 MOS of 4.32), with a real-time factor of only 0.12 and fewer than 5 million parameters. It significantly outperforms existing generative approaches while maintaining high intelligibility and naturalness. The method thus offers an unprecedented combination of efficiency, lightweight design, and practical deployability.

Technology Category

• Machine Learning: Deep Generative Models & Autoencoders
• Natural Language Processing: Generation
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Large language models for search
• Security and Privacy: Privacy-enhancing technologies

Abstract

Speech enhancement (SE) recovers clean speech from noisy signals and is vital for applications such as telecommunications and automatic speech recognition (ASR). While generative approaches achieve strong perceptual quality, they often rely on multi-step sampling (diffusion/flow-matching) or large language models, limiting real-time deployment. To mitigate these constraints, we present MeanFlowSE, a one-step generative SE framework. It adopts MeanFlow to predict an average-velocity field for one-step latent refinement and conditions the model on self-supervised learning (SSL) representations rather than VAE latents. This design accelerates inference and provides robust acoustic-semantic guidance during training. In the Interspeech 2020 DNS Challenge blind test set and simulated test set, MeanFlowSE attains state-of-the-art (SOTA) level perceptual quality and competitive intelligibility while significantly lowering both real-time factor (RTF) and model size compared with recent generative competitors, making it suitable for practical use. The code will be released upon publication at https://github.com/Hello3orld/MeanFlowSE.

Problem

• One-step generative speech enhancement for real-time deployment
• Improving perceptual quality and intelligibility of noisy speech
• Reducing computational cost and model size for practicality

Innovation

• One-step generative speech enhancement via MeanFlow
• Uses self-supervised learning representations for conditioning
• Achieves real-time efficiency with reduced model size