Fault Detection of In-Service Bridge Expansion Joint Based on Voiceprint Recognition

Table 5

Performance comparison of event segmentation models.

Item	Unit	ML models		Threshold models
		ANN	SVM	STE	STCZ
Optimized hyperparameter	—	35	0.5	8	6
Accuracy	%	93.8	93.1	90.0	51.8
AuC	—	0.981	0.964	—	—
Time of training	s	36.7	3613.9	0	0
Time of inference	ms/s	10.5	242.9	208.1	250.5