Fig. 6From: Improved Wasserstein conditional generative adversarial network speech enhancementThe SNR of enhancement speech. It depicts the SNR of speech enhancement signal of test data in improved WCGAN models with different numbers of scenes and SEGAN model with the average SNR. We find surprisingly that the improved model achieves mean value 23.65 gain in the SNR on WCGAN and achieves 21.05 gain on SEGANBack to article page