Skip to main content

Table 8 Mean objective results with 75% overlap for factory noise

From: An adaptive a priori SNR estimator for perceptual speech enhancement

Gain SNR PESQ SNR seg HD KurtR
   DD MDD TSNR Prop DD MDD TSNR Prop DD MDD TSNR Prop DD MDD TSNR Prop
WF 0 2.356 2.354 2.341 2 . 3 9 8 1.201 1.260 1.358 1 . 6 5 2 0.357 0.375 0.355 0 . 3 4 8 1.075 1.064 1 . 0 6 2 1 1.081
  5 2.693 2.687 2.670 2 . 7 3 8 4.035 4.096 4.182 4 . 5 0 4 0.339 0.340 0.339 0 . 3 3 0 1.178 1.138 1 . 1 3 7 1.158
  10 3.099 3.091 3.066 3 . 1 1 8 7.089 7.159 7.250 7 . 6 5 5 0.314 0.317 0.326 0 . 3 0 0 1.483 1.340 1 . 3 2 7 1.435
LSA 0 2.392 2.406 2.238 2 . 4 5 2 1.169 1.374 1.067 1 . 6 7 4 0.345 0.352 0.362 0 . 3 4 4 1.153 1 . 0 7 9 1.383 1.112
  5 2.722 2.749 2.641 2 . 7 9 9 4.029 4.237 3.970 4 . 5 5 6 0.323 0.332 0.315 0 . 3 2 0 1.401 1 . 1 6 5 1.677 1.231
  10 3.067 3.144 3.023 3 . 1 6 9 7.142 7.335 7.149 7 . 7 4 7 0 . 2 8 7 0.305 0.297 0.296 1.917 1 . 4 0 2 1950 1.507
  1. Bold values denote the best performance