Comparative Study between Adversarial Networks and Classical Techniques for Speech Enhancement
This work provides a comparative analysis for researchers and practitioners in speech enhancement, though it is incremental as it evaluates existing methods without introducing new ones.
This study compared classical speech enhancement techniques (Wiener filter, LogMMSE) with a deep learning approach (SEGAN) across 85 noise conditions, finding that classical methods generally performed better but SEGAN excelled in severe noise scenarios with lower variance.
Speech enhancement is a crucial task for several applications. Among the most explored techniques are the Wiener filter and the LogMMSE, but approaches exploring deep learning adapted to this task, such as SEGAN, have presented relevant results. This study compared the performance of the mentioned techniques in 85 noise conditions regarding quality, intelligibility, and distortion; and concluded that classical techniques continue to exhibit superior results for most scenarios, but, in severe noise scenarios, SEGAN performed better and with lower variance.