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.

翻译：尽管多通道语音分离的判别模型在参考指标上表现优异，但其人类听觉质量往往不尽如人意。为解决这一问题，我们提出了一种基于均值流的一步式生成校正器（MeCo）。MeCo通过学习条件平均速度场，将判别估计值在单步内直接映射到干净语音流形上。为最大化一步式生成性能，我们引入了数据空间优化（DSO）。DSO整合了$\mathbf{x}_r$-损失（该损失惩罚较长位移区间上的预测误差，作为提升人类听觉质量的生成目标）与端点SI-SDR损失（直接优化终端信号保真度）。实验表明，MeCo在最小计算开销下实现了最先进（SOTA）性能，同时在域内和域外场景中均获得了卓越的信号保真度与人类听觉质量。