Comparative analysis of generative versus discriminative speech enhancement models shows differences in robustness to noise, model complexity, convergence, and hallucination measured via word error rate and phoneme similarity.
DeepFilterGAN: A full-band real-time speech enhancement system with GAN-based stochastic regeneration,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
eess.AS 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
A Comparison of Generative and Discriminative Methods for Speech Enhancement: Robustness, Complexity, and Hallucination
Comparative analysis of generative versus discriminative speech enhancement models shows differences in robustness to noise, model complexity, convergence, and hallucination measured via word error rate and phoneme similarity.