A Comprehensive Review on Generative Models for Speech Enhancement
Nidula Elgiriyewithana, Nihal D. Kodikara · 2024
This review paper provides an exhaustive analysis of the current state of generative models used for speech enhancement, with a focus on Autoencoders, Diffusion models, and Generative Adver- sarial Networks (GANs). Particular emphasis is placed on understanding their principles, architecture, and applicability to speech enhancement. Furthermore, we delve into the advantages and limitations of these models, highlighting methodological progression, and recent trends. The study also emphasizes the significance of Public Perception of Current Speech Enhancement Methods and discusses potential future directions and challenges in the field - exploring strategies for improving generalization, robustness, and the incorporation of multi-modal information. The conclusion summarizes the key findings of the collected literature, suggesting potential applications and impacts of the discussed generative models for speech enhancement.