Skip to main content

Table 5 Performance of the proposed and other discrimination algorithms

From: Music detection from broadcast contents using convolutional neural networks with a Mel-scale kernel

Test data Algorithm F-score (%) Precision (%) Recall (%) Recall_nosp (%)
Korean drama Proposed 95.9 95.9 96.0 96.8
Tsipas et al. [16] 77.9 97.0 65.1 74.6
Doukhan et al. [15] 79.9 95.4 68.7 80.3
Korean reality Proposed 94.7 93.0 96.4 97.5
Tsipas et al. [16] 66.6 96.2 51.0 67.6
Doukhan et al. [15] 68.0 96.4 52.5 75.6
British 8 h Proposed 86.5 85.3 87.8 92.2
Tsipas et al. [16] 67.8 83.2 57.3 72.2
Doukhan et al. [15] 67.4 80.6 57.9 86.7
Spanish 12 h Proposed 88.9 84.7 93.4 96.8
Tsipas et al. [16] 71.1 91.8 58.0 76.4
Doukhan et al. [15] 73.1 92.9 60.2 86.5
MIREX 2015 Proposed 95.3 99.4 91.6 92.5
Tsipas et al. [16] 96.4 99.3 93.7 96.1
Doukhan et al. [15] 95.4 99.3 91.9 94.1
Dafx 07 Proposed 87.6 88.3 87.0 87.0
Tsipas et al. [16] 70.6 88.7 58.7 58.7
Doukhan et al. [15] 65.7 87.9 52.4 52.4