AlphaFlowTSE: One-Step Generative Target Speaker Extraction via Conditional AlphaFlow
This work addresses latency and reliability issues in real-world conversations for speech processing applications, representing an incremental improvement over existing methods.
The paper tackled the problem of target speaker extraction from multi-talker mixtures by proposing AlphaFlowTSE, a one-step generative model that improves target-speaker similarity and generalization for automatic speech recognition, achieving better performance on Libri2Mix and REAL-T datasets.
In target speaker extraction (TSE), we aim to recover target speech from a multi-talker mixture using a short enrollment utterance as reference. Recent studies on diffusion and flow-matching generators have improved target-speech fidelity. However, multi-step sampling increases latency, and one-step solutions often rely on a mixture-dependent time coordinate that can be unreliable for real-world conversations. We present AlphaFlowTSE, a one-step conditional generative model trained with a Jacobian-vector product (JVP)-free AlphaFlow objective. AlphaFlowTSE learns mean-velocity transport along a mixture-to-target trajectory starting from the observed mixture, eliminating auxiliary mixing-ratio prediction, and stabilizes training by combining flow matching with an interval-consistency teacher-student target. Experiments on Libri2Mix and REAL-T confirm that AlphaFlowTSE improves target-speaker similarity and real-mixture generalization for downstream automatic speech recognition (ASR).