Skip to main content
Figure 3 | SpringerPlus

Figure 3

From: On pre-image iterations for speech enhancement

Figure 3

PI in a speech segment shown on the magnitude of the spectral data of the utterance in Figure 8. The columns show from left to right: (i) Noisy segment with one frequency band and a patch marked. The noise level is 10 dB SNR. (ii) Enhanced segment. (iii) The marked frequency band before de-noising and after one to three iterations. (iv) Kernel values between the marked patch and all other noisy patches in the band before de-noising and after one to three iterations. The kernel vector contains only one value significantly larger than zero and no averaging is performed for this patch. Note that the patches are extracted row-wise from left to right and from top to bottom.

Back to article page