aiqbench analysis
DriftSE introduces a novel generative framework that reformulates speech denoising as an equilibrium problem to achieve high-fidelity enhancement in a single inference step. By utilizing a learned Drifting Field to match distributions rather than relying on iterative sampling, the method effectively handles unpaired data and outperforms existing multi-step diffusion baselines on the VoiceBank-DEMAND benchmark. This approach establishes a new paradigm for audio processing that eliminates the latency typically associated with diffusion-based models.
Speech Enhancement Based on Drifting Models
arXiv:2604.24199v4 Announce Type: replace-cross Abstract: We propose Speech Enhancement based on Drifting Models (DriftSE), a novel generative framework that formulates denoising as an equilibrium problem. Rather than relying on iterative sampling, DriftSE natively achieves one-step inference by evolving the pushforward distribution of a mapping function to directly match the clean speech distribution. This evolution is driven by a Drifting Field, a learned correction vector that guides samples toward the high-density regions of the clean distribution, which naturally facilitates training on unpaired data by matching distributions rather than paired samples. We investigate the framework under two formulations: a direct mapping from the noisy observation, and a stochastic conditional generative model from a Gaussian prior. Experiments on the VoiceBank-DEMAND benchmark demonstrate that DriftSE achieves high-fidelity enhancement in a single step, outperforming multi-step diffusion baselines and establishing a new paradigm for speech enhancement.
Abstract:We propose Speech Enhancement based on Drifting Models (DriftSE), a novel generative framework that formulates denoising as an equilibrium problem. Rather than relying on iterative sampling, DriftSE natively achieves one-step inference by evolving the pushforward distribution of a mapping function to directly match the clean speech distribution. This evolution is driven by a Drifting Field, a learned correction vector that guides samples toward the high-density regions of the clean distribution, which naturally facilitates training on unpaired data by matching distributions rather than paired samples. We investigate the framework under two formulations: a direct mapping from the noisy observation, and a stochastic conditional generative model from a Gaussian prior. Experiments on the VoiceBank-DEMAND benchmark demonstrate that DriftSE achieves high-fidelity enhancement in a single step, outperforming multi-step diffusion baselines and establishing a new paradigm for speech enhancement.
Sound (cs.SD); Artificial Intelligence (cs.AI); Audio and Speech Processing (eess.AS); Signal Processing (eess.SP)
arXiv:2604.24199 [cs.SD]
(or arXiv:2604.24199v4 [cs.SD] for this version)
Full text under the publisher's open licence · 156 words
Related coverage
Regularized Schr\"odinger Bridge via Distortion-Perception Perturbation for High-Fidelity Speech Enhancement
arXiv · Aug 4, 2026
Honest Abacus AI Review: ChatLLM, DeepAgent, AI Studio & More
KDnuggets · Aug 4, 2026
Hi-Singers: A Comprehensive High-Quality Dataset for Expressive Audio-Driven Singing Head Synthesis
arXiv · Sep 22, 2026
Assessing Information Ecosystems : How Governments Can Get Ahead of Hybrid Threats
carnegieendowment.org · Aug 5, 2026