EURASIP Journal on Audio, Speech, and Music Processing

Table 4 Model performance comparison under different attention mechanisms based on FOA and MIC format dataset

From: Attention mechanism combined with residual recurrent neural network for sound event detection and localization

The model name	On FOA format dataset				On MIC format dataset
	DE (\(^\circ\))	FR (\(\%\))	ER	F1-score (\(\%\))	DE (\(^\circ\))	FR (\(\%\))	ER	F1-score (\(\%\))
sSE	7.6	88.4	0.24	87.8	8.0	87.4	0.26	86.8
cSE	7.5	89.3	0.25	87.7	7.9	88.3	0.28	86.7
scSE	5.1	89.6	0.21	88.4	5.5	88.5	0.23	87.4
Not using attention	13.5	84.7	0.28	82.1	14.1	83.4	0.31	83.1

Back to article page