To address the existing shortcomings of the real valued convolutional neural network based estimation of vector hydrophone direction of arrival(DOA) that does not adequately extract the phase features of the signals received by the array,a vector hydrophone DOA estimation method(CV-CDAE-CNN) based on combination of complex-valued convolutional denoising autoencoder(CV-CDAE) and complex-valued convolutional neural network(CV-CNN) is proposed. The complex-valued covariance matrix of the received signals from the vector hydrophone are firstly input into the CV-CDAE module to remove the noise,and then the denoised samples are input into the CV-CNN for classification. The double-scale dilation convolution is used to increase the receptive field of the feature map of CV-CNN before down-sampling and to mitigate the information loss caused by down-sampling. Angle classification is realized by CV-CDAE denoising and CV-CNN's unique way of dealing with complex values, and then DOA estimates are obtained. Simulation results show that the proposed method in this paper has stronger generalization ability, higher DOA estimation accuracy and higher estimation accuracy compared with the existing CV-CNN at low signal-to-noise ratio or limited number of snapshots.
NIUR, ZHANGZ. Harmonic frequency enhanced direction of arrival estimation using non-uniform linear arrays[J]. IEEE Access, 2024, 12: 36942-36950.
[2]
YUJ, WANGY. Deep learning-based multipath DoAs estimation method for mmWave massive MIMO systems in low SNR[J]. IEEE Transactions on Vehicular Technology, 2023, 72(6): 7480-7490.
[3]
PANM, LIUP, LIUS, et al. Efficient joint DOA and TOA estimation for indoor positioning with 5G picocell base stations[J]. IEEE Transactions on Instrumentation and Measurement, 2022, 71: 8005219.
[4]
XUZ, LIH, YANGK. A modified differential beamforming and its application for DOA estimation of low frequency underwater signal[J]. IEEE Sensors Journal, 2020, 20(16): 8890-8902.
[5]
SCHMIDTR. Multiple emitter location and signal parameter estimation[J]. IEEE Transactions on Antennas and Propagation, 1986, 34(3): 276-280.
[6]
ROYR, PAULRAJA, KAILATHT. ESPRIT: a subspace rotation approach to estimation of parameters of cisoids in noise[J]. IEEE Transactions on Acoustics, Speech, and Signal Processing, 1986, 34(5): 1340-1342.
[7]
MERKOFERJ P, REVACHG, SHLEZINGERN, et al. DA-MUSIC: data-driven DoA estimation via deep augmented MUSIC algorithm[J]. IEEE Transactions on Vehicular Technology, 2024, 73(2): 2771-2785.
[8]
NINGY M, MAS, MENGF Y, et al. DOA estimation based on ESPRIT algorithm method for frequency scanning LWA[J]. IEEE Communications Letters, 2020, 24(7): 1441-1445.
[9]
LIW, ZHUZ, GAOW, et al. Stability and super-resolution of MUSIC and ESPRIT for multi-snapshot spectral estimation[J]. IEEE Transactions on Signal Processing, 2022, 70: 4555-4570.
[10]
WANGP, HEW, GUOF, et al. An improved atomic search algorithm for optimization and application in ML DOA estimation of vector hydrophone array[J]. AIMS Mathematics, 2022, 7(4): 5563-5593.
[11]
GERSTOFTP, MECKLENBRÄUKERC F, XENAKIA, et al. Multisnapshot sparse Bayesian learning for DOA[J]. IEEE Signal Processing Letters, 2016, 23(10): 1469-1473.
[12]
KATTENBORNT, LEITLOFFJ, SCHIEFERF, et al. Review on Convolutional Neural Networks (CNN) in vegetation remote sensing[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2021, 173: 24-49.
[13]
YUANF, ZHANGZ, FANGZ. An effective CNN and Transformer complementary network for medical image segmentation[J]. Pattern Recognition, 2023, 136: 109228.
[14]
YANGR, ZHAX, LIUK, et al. A CNN model embedded with local feature knowledge and its application to time-varying signal classification[J]. Neural Networks, 2021, 142: 564-572.
[15]
LIUY, CHENH, WANGB. DOA estimation based on CNN for underwater acoustic array[J]. Applied Acoustics, 2021, 172: 107594.
[16]
PAPAGEORGIOUG K, SELLATHURAIM, ELDARY C. Deep networks for direction-of-arrival estimation in low SNR[J]. IEEE Transactions on Signal Processing, 2021, 69: 3714-3729.
[17]
WUL, LIUZ M, HUANGZ T. Deep convolution network for direction of arrival estimation with sparse prior[J]. IEEE Signal Processing Letters, 2019, 26(11): 1688-1692.
[18]
STROGATZS H. Exploring complex networks[J]. Nature, 2001, 410(6825): 268-276.
YUFan, CHENGege, SHENMingwei. DOA estimation algorithm based on dual-channel complex-valued convolutional neural network[J]. Modern Radar, 2022, 44(12): 81-86. (in Chinese)
[21]
CHENY, HEH, LIUS, et al. Physical layer authentication for industrial control based on convolutional denoising autoencoder[J]. IEEE Internet of Things Journal, 2024, 11(9): 15633-15641.
[22]
WANGJ, XIEX, SHIJ, et al. Denoising autoencoder, A deep learning algorithm, aids the identification of a novel molecular signature of lung adenocarcinoma[J]. Genomics, Proteomics & Bioinformatics, 2020, 18(4): 468-480.
[23]
BANKD, KOENIGSTEINN, GIRYESR. Autoencoder[M]//Machine Learning for Data Science Handbook. Cham: Springer International Publishing, 2023: 353-374.