aiqbench analysis
MeCo addresses the gap between high reference-based metrics and poor human listening quality in multi-channel speech separation by introducing a one-step generative corrector. By leveraging MeanFlow to map estimates directly onto the clean speech manifold and optimizing via Data-Space Optimization, the method achieves state-of-the-art fidelity with minimal computational overhead. This approach offers a practical pathway for improving audio processing pipelines where perceptual quality is critical alongside signal accuracy.
MeCo: One-Step MeanFlow-based Corrector for Multi-Channel Speech Separation
arXiv:2606.09677v1 Announce Type: cross Abstract: While discriminative models for multi-channel speech separation excel in reference-based metrics, they often exhibit suboptimal human listening quality. To address this, we propose a novel MeanFlow-based one-step generative corrector (MeCo). MeCo learns a conditional average velocity field to map discriminative estimates directly onto the clean speech manifold in a single step. To maximize one-step generation performance, we introduce Data-Space Optimization (DSO). DSO integrates an $\mathbf{x}_r$-loss, which penalizes prediction errors on longer displacement intervals to serve as a generative objective for human listening quality, with an Endpoint SI-SDR loss that directly optimizes terminal signal fidelity. Experiments demonstrate that MeCo achieves state-of-the-art (SOTA) performance with minimal computational overhead, simultaneously achieving superior signal fidelity and human listening quality in both in-domain and out-of-domain scenarios.
Abstract:While discriminative models for multi-channel speech separation excel in reference-based metrics, they often exhibit suboptimal human listening quality. To address this, we propose a novel MeanFlow-based one-step generative corrector (MeCo). MeCo learns a conditional average velocity field to map discriminative estimates directly onto the clean speech manifold in a single step. To maximize one-step generation performance, we introduce Data-Space Optimization (DSO). DSO integrates an $\mathbf{x}_r$-loss, which penalizes prediction errors on longer displacement intervals to serve as a generative objective for human listening quality, with an Endpoint SI-SDR loss that directly optimizes terminal signal fidelity. Experiments demonstrate that MeCo achieves state-of-the-art (SOTA) performance with minimal computational overhead, simultaneously achieving superior signal fidelity and human listening quality in both in-domain and out-of-domain scenarios.
Audio and Speech Processing (eess.AS); Artificial Intelligence (cs.AI)
arXiv:2606.09677 [eess.AS]
(or arXiv:2606.09677v1 [eess.AS] for this version)
https://doi.org/10.48550/arXiv.2606.09677 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 220 words
Related coverage
MeCo: One-Step MeanFlow-based Corrector for Multi-Channel Speech Separation
arXiv · Jun 10, 2026
Models Can't Remember Their Training. Neither Can You.
LessWrong · Jul 19, 2026
Six Principles of Voice Design: Designing for the Ear and Encoding It into Agent Script
Salesforce Blog · Oct 8, 2026
2026.9: There's room on this bus
Home Assistant · Sep 2, 2026