Skip to main content
. Author manuscript; available in PMC: 2022 Jan 1.
Published in final edited form as: IEEE/ACM Trans Audio Speech Lang Process. 2021 Jan 25;29:2058–2066. doi: 10.1109/taslp.2021.3054302

TABLE VII.

Experimental Results of Different Downsampling Schemes Evaluated on TIMIT

Subsampling Decimating FFT
Model SNR LSD PESQ SNR LSD PESQ SNR LSD PESQ
Trained with Subsampling 20.18 0.72 3.65 15.32 1.28 3.15 16.02 0.88 3.52
Trained with Decimating −10.85 2.51 1.03 17.94 0.75 3.95 −12.33 2.48 1.03
Trained with FFT 14.55 0.84 2.59 15.62 1.35 3.32 17.28 0.72 3.92
Trained with Random 19.57 0.75 3.51 17.07 0.77 3.81 16.74 0.77 3.88