Abstract
Training target speaker extraction (TSE) models for real conversational mixtures remains challenging due to the unavailability of large-scale training corpora and clean target speech for supervision. We present PS4, a proxy-supervised training framework for TSE in real conversational mixtures, with two main contributions.
First, we construct a large-scale corpus of 71,771 training samples derived from four public datasets, covering both Chinese and English scenarios. Each sample contains an overlapping speech mixture, per-speaker enrollment audio, a ground-truth transcript, and frame-level voice activity labels.
Second, we propose a proxy-supervised joint training strategy that fine-tunes a BSRNN-based TSE model using four complementary differentiable objectives: ASR cross-entropy, speaker similarity, frame-level voice activity detection, and perceptual audio quality. Starting from a publicly available pre-trained checkpoint, only the BSRNN separator is updated during fine-tuning. On the REAL-T challenge leaderboard, PS4 ranks 2nd overall, achieving the best speaker similarity and timing F1 among all submitted systems.
Blogger's Review: The PS4 framework effectively addresses the challenges of target speaker extraction in real-world scenarios through proxy supervision, demonstrating strong potential, especially in data-scarce situations. The constructed large-scale corpus and multi-objective joint training strategy provide new insights for future research. Its impressive performance in the REAL-T challenge further validates the method's effectiveness.