Figure 6

Comparison of one randomly selected piece of speech data from the M-DGAN test set. (a) denotes clean speech, (b) denotes noisy speech, (c) denotes the output of the TF-___domain two-way speech enhancement structure, (d) denotes the output of the T-___domain residual noise estimation structure, and (e) denotes the final enhanced speech generated by mixing (c,d).