Skip to main content

Advertisement

Table 1 AUC values of the evaluated VAD algorithms under −5 dB SNR

From: Voice activity detection algorithm based on long-term pitch information

Noise Sohn Harmfreq LTSD LTSV LSFM LTPD
factory1 0.5538 0.5542 0.5978 0.8223 0.7113 0.8998
factory2 0.8702 0.8678 0.8508 0.9139 0.9190 0.9266
leopard 0.9608 0.9608 0.8721 0.9555 0.9623 0.9435
m109 0.9182 0.9096 0.8787 0.9653 0.9587 0.9500
opsroom 0.8183 0.8065 0.8498 0.9103 0.8561 0.8696
f16 0.8614 0.8587 0.8794 0.9296 0.8997 0.9316
buccaneer1 0.7612 0.7505 0.8471 0.9163 0.7921 0.9382
buccaneer2 0.8162 0.8119 0.8794 0.9494 0.9086 0.9495
babble 0.7687 0.7676 0.8556 0.7310 0.6873 0.7788
engine 0.8556 0.8521 0.9036 0.9546 0.8791 0.9069
hfchannel 0.8814 0.8797 0.9134 0.9480 0.8626 0.9312
machinegun 0.5934 0.5869 0.7860 0.7481 0.3423 0.9380
pink 0.7802 0.7776 0.8609 0.9434 0.8777 0.9481
volvo 0.9594 0.9594 0.9327 0.9273 0.9501 0.9561
white 0.8601 0.8572 0.8901 0.9609 0.9096 0.9521
average 0.8172 0.8134 0.8532 0.9051 0.8352 0.9213
  1. Note: The italicized numbers mean the best performance among all evaluated algorithms with the specific noise