Audio-Visual Speech Enhancement with Score-Based Generative Models

Inhalt:
This paper introduces an audio-visual speech enhancement system that leverages score-based generative models, also known as diffusion models, conditioned on visual information. In particular, we exploit audio-visual embeddings obtained from a self-supervised learning model that has been fine-tuned on lipreading. The layer-wise features of its transformer-based encoder are aggregated, time-aligned, and incorporated into the noise conditional score network. Experimental evaluations show that the proposed audiovisual speech enhancement system yields improved speech quality and reduces generative artifacts such as phonetic confusions with respect to the audio-only equivalent. The latter is supported by the word error rate of a downstream automatic speech recognition model, which decreases noticeably, especially at low input signal-to-noise ratios.