Language agnostic missing subtitle detection

EURASIP Journal on Audio, Speech, and Music Processing

Table 6 Block-level analysis on human annotated dataset. ↓: lower is better and ↑: higher is better

Threshold (t in ms)	FPR ↓			Precision ↑			Recall ↑
	VAD	VAD +AC	% Reduction	VAD	VAD + AC	% Improvement	VAD	VAD + AC	% Improvement
300	0.214	0.106	50.467	0.277	0.438	58.123	0.712	0.69	− 3.090
500	0.195	0.072	63.077	0.298	0.552	85.235	0.707	0.667	− 5.658
800	0.169	0.039	76.923	0.334	0.732	119.162	0.702	0.631	− 10.114
1000	0.149	0.03	79.866	0.369	0.819	121.951	0.695	0.597	− 14.101
1200	0.134	0.024	82.090	0.398	0.853	114.322	0.687	0.558	− 18.777