DiffPhase: Generative Diffusion-based STFT Phase Retrieval
This work addresses phase retrieval for speech processing, offering an incremental improvement by applying an existing diffusion method to a specific domain.
The authors tackled the problem of STFT phase retrieval by adapting a diffusion model from speech enhancement, achieving performance that surpasses both classical and modern methods in speech quality and intelligibility metrics.
Diffusion probabilistic models have been recently used in a variety of tasks, including speech enhancement and synthesis. As a generative approach, diffusion models have been shown to be especially suitable for imputation problems, where missing data is generated based on existing data. Phase retrieval is inherently an imputation problem, where phase information has to be generated based on the given magnitude. In this work we build upon previous work in the speech domain, adapting a speech enhancement diffusion model specifically for STFT phase retrieval. Evaluation using speech quality and intelligibility metrics shows the diffusion approach is well-suited to the phase retrieval task, with performance surpassing both classical and modern methods.