TY - GEN
T1 - Underwater acoustic target recognition based on u-shaped network
AU - Lingzhi, Xue
AU - Xiangyang, Zeng
N1 - Publisher Copyright:
© INTER-NOISE 2021 .All right reserved.
PY - 2021
Y1 - 2021
N2 - Underwater target recognition is an essential but difficult technique in acoustic signal processing. The most serious challenge of the underwater recognition is the scarcity of underwater acoustic samples. To solve this problem, this paper proposed a local skip connection U-shaped architecture network (U-Net), which is based on the convolutional neural network (CNN). The network architecture is designed ingeniously with generating a contracting path and an expansive path to achieve the extraction of different scale features, so as to improve the classification rate. The experimental results based on the measured data demonstrate that the recognition accuracy of our proposed scheme performs better than that of the deep belief network (DBN), deep auto-encoder (DAE) model and UATC-Densenet (underwater acoustic target classification Densenet) model. Meanwhile, the visualization of these four networks shows that the proposed network can learn more effective feature information with limited samples.
AB - Underwater target recognition is an essential but difficult technique in acoustic signal processing. The most serious challenge of the underwater recognition is the scarcity of underwater acoustic samples. To solve this problem, this paper proposed a local skip connection U-shaped architecture network (U-Net), which is based on the convolutional neural network (CNN). The network architecture is designed ingeniously with generating a contracting path and an expansive path to achieve the extraction of different scale features, so as to improve the classification rate. The experimental results based on the measured data demonstrate that the recognition accuracy of our proposed scheme performs better than that of the deep belief network (DBN), deep auto-encoder (DAE) model and UATC-Densenet (underwater acoustic target classification Densenet) model. Meanwhile, the visualization of these four networks shows that the proposed network can learn more effective feature information with limited samples.
UR - http://www.scopus.com/inward/record.url?scp=85117404655&partnerID=8YFLogxK
U2 - 10.3397/IN-2021-1820
DO - 10.3397/IN-2021-1820
M3 - 会议稿件
AN - SCOPUS:85117404655
T3 - Proceedings of INTER-NOISE 2021 - 2021 International Congress and Exposition of Noise Control Engineering
BT - Proceedings of INTER-NOISE 2021 - 2021 International Congress and Exposition of Noise Control Engineering
A2 - Dare, Tyler
A2 - Bolton, Stuart
A2 - Davies, Patricia
A2 - Xue, Yutong
A2 - Ebbitt, Gordon
PB - The Institute of Noise Control Engineering of the USA, Inc.
T2 - 50th International Congress and Exposition of Noise Control Engineering, INTER-NOISE 2021
Y2 - 1 August 2021 through 5 August 2021
ER -