Skip to main content
. 2020 Sep 8;20(18):5112. doi: 10.3390/s20185112

Table 15.

F1 scores of the models considered in this work (as a percentage) for noise rejection performance (best values are highlighted in bold).

SNR RNN-WDCNN WDCNN SRDCNN FFT-SVM FFT-MLP
Average 99.79 (± 0.16) 93.63 (± 1.98) 73.29 (± 3.79) 90.71 (± 0.36) 89.62 (± 0.38)
None 100.00 (± 0.00) 100.00 (± 0.01) 100.00 (± 0.00) 99.63 (± 0.08) 99.51 (± 0.10)
10 dB 100.00 (± 0.00) 100.00 (± 0.01) 99.88 (± 0.06) 99.26 (± 0.13) 98.09 (± 0.23)
8 dB 100.00 (± 0.00) 99.99 (± 0.01) 99.67 (± 0.17) 98.68 (± 0.19) 97.91 (± 0.12)
6 dB 100.00 (± 0.00) 99.99 (± 0.02) 98.14 (± 1.61) 98.21 (± 0.31) 97.08 (± 0.31)
4 dB 100.00 (± 0.00) 99.90 (± 0.07) 89.65 (± 6.64) 96.10 (± 0.38) 95.37 (± 0.43)
2 dB 100.00 (± 0.00) 98.81 (± 0.63) 73.43 (± 6.41) 93.93 (± 0.38) 92.95 (± 0.45)
0 dB 99.96 (± 0.06) 93.67 (± 2.92) 54.68 (± 6.94) 87.29 (± 0.51) 84.86 (± 0.72)
−2 dB 99.67 (± 0.25) 82.66 (± 6.18) 30.03 (± 6.42) 76.46 (± 0.70) 75.91 (± 0.62)
−4 dB 98.45 (± 1.11) 67.70 (± 7.94) 14.12 (± 5.84) 66.84 (± 0.59) 64.89 (± 0.47)