A Closer Look at Wav2Vec2 Embeddings for On-Device Single-Channel Speech Enhancement
This work addresses the problem of efficient speech enhancement for on-device applications, but it is incremental as it primarily evaluates existing methods rather than introducing new ones.
The paper investigated the utility of self-supervised learned (SSL) representations, specifically Wav2Vec2 embeddings, for on-device single-channel speech enhancement in low SNR conditions, finding that they add very little value to the enhancement task.
Self-supervised learned models have been found to be very effective for certain speech tasks such as automatic speech recognition, speaker identification, keyword spotting and others. While the features are undeniably useful in speech recognition and associated tasks, their utility in speech enhancement systems is yet to be firmly established, and perhaps not properly understood. In this paper, we investigate the uses of SSL representations for single-channel speech enhancement in challenging conditions and find that they add very little value for the enhancement task. Our constraints are designed around on-device real-time speech enhancement -- model is causal, the compute footprint is small. Additionally, we focus on low SNR conditions where such models struggle to provide good enhancement. In order to systematically examine how SSL representations impact performance of such enhancement models, we propose a variety of techniques to utilize these embeddings which include different forms of knowledge-distillation and pre-training.