Deep Learning Methods for Underwater Target Feature Extraction and Recognition

Table 4

Comparison of classification effects of different features.

Feature types	Number of hidden layer neurons	ELM activated functions	Recognition rates (%)
MFCC	40	Sigmoid	84.64
	60	Sigmoid	84.48
	80	Sigmoid	84.41
	60	tanh	82.39
HHT	40	Sigmoid	81.06
	60	Sigmoid	82.34
	40	tanh	81.72
	60	tanh	82.04
Deep learning feature	40	Sigmoid	90.39
	60	Sigmoid	92.69
	20	tanh	92.40
	40	tanh	93.04
	60	tanh	92.29