DriftSE: Speech Enhancement with Generative Drifting
This work introduces dual-latent drifting, performing parallel drifting in both semantic and acoustic latents to simultaneously preserve phonetic intelligibility and acoustic fidelity and demonstrates that DriftSE enables fully unpaired training by aligning latent distributions rather than exact point-wise targets.
Liang Xu, Diego Caviedes-Nozal, W. Kleijn et al.
· 1 citation