表 1. Average results of automatic detection of arousal events based on the convolutional self-attention model and its variants(
).
基于卷积注意力模型及其变体的觉醒事件自动检测平均结果()
网络模型 | 准确率(%) | F1分数(%) | AUPRC(%) | AUROC(%) |
卷积注意力模型 | 81.9 ± 5.0 | 74.9 ± 12.6 | 68.1 ± 17.2 | 78.9 ± 6.5 |
30 s-卷积注意力模型 | 81.9 ± 8.0 | 64.3 ± 13.3 | 55.7 ± 17.2 | 75.0 ± 7.0 |
基线模型 | 74.2 ± 9.0 | 65.8 ± 13.9 | 60.7 ± 18.9 | 72.3 ± 7.8 |
单一尺度卷积注意力模型 | 79.2 ± 6.7 | 71.3 ± 12.2 | 65.2 ± 17.7 | 76.5 ± 6.4 |
无SA卷积注意力模型 | 78.3 ± 8.0 | 70.4 ± 14.2 | 63.9 ± 18.7 | 75.3 ± 7.3 |