Bridging Self-Supervised Learning and Speech Enhancement: A Wav2Vec2-Conditioned Framework | ArxivCSExplorer