GenTSE: Enhancing Target Speaker Extraction via a Coarse-to-Fine Generative Language Model
This work addresses speech separation for applications like hearing aids or communication systems, but it is incremental as it builds on existing LM-based methods with specific enhancements.
The paper tackles target speaker extraction by introducing GenTSE, a two-stage generative language model that separates coarse semantic and fine acoustic token generation, resulting in improved speech quality, intelligibility, and speaker consistency on Libri2Mix benchmarks.
Language Model (LM)-based generative modeling has emerged as a promising direction for TSE, offering potential for improved generalization and high-fidelity speech. We present GenTSE, a two-stage decoder-only generative LM approach for TSE: Stage-1 predicts coarse semantic tokens, and Stage-2 generates fine acoustic tokens. Separating semantics and acoustics stabilizes decoding and yields more faithful, content-aligned target speech. Both stages use continuous SSL or codec embeddings, offering richer context than discretized-prompt methods. To reduce exposure bias, we employ a Frozen-LM Conditioning training strategy that conditions the LMs on predicted tokens from earlier checkpoints to reduce the gap between teacher-forcing training and autoregressive inference. We further employ DPO to better align outputs with human perceptual preferences. Experiments on Libri2Mix show that GenTSE surpasses previous LM-based systems in speech quality, intelligibility, and speaker consistency.