TY - GEN
T1 - Instantaneous Hybrid Signal Separation Based on CANDECOMP/PARAFAC Decomposition with Accelerated Proximal Gradient Method
AU - Wu, Han
AU - He, Shengyang
AU - Ren, Guanghui
AU - Yang, Rongqian
AU - Zhao, Yaqin
AU - Wu, Longwen
N1 - Publisher Copyright:
© 2022 IEEE.
PY - 2022
Y1 - 2022
N2 - In this paper, to solve the anti-interference problem of UAVs in bland source decomposition, a new signal separation method of instantaneous hybrid model based on the classic method named CANDECOMP/PARAFAC Decomposition (CPD) is studied. Firstly, the instantaneous hybrid signal decomposition is analyzed with an instantaneous hybrid model and a regular parallel decomposition method. Secondly, the signal separation problem applied in the instantaneous hybrid model is solved by using a regular parallel decomposition method from tensor decomposition theory. Finally, motivated by the idea of algorithm optimization by adding an operator, a new CPD method with proximal gradient constraint based on the classical CPD separation algorithms is proposed, which can be called as CPD-APG. Compared with the conventional JADE separation method and FastICA separation method, the separation error of the proposed algorithm is generally reduced by about 50% under the SNR range of-20dB~0dB. When the SNR is above 0dB, the separation error of the proposed algorithm is nearly close to zero.
AB - In this paper, to solve the anti-interference problem of UAVs in bland source decomposition, a new signal separation method of instantaneous hybrid model based on the classic method named CANDECOMP/PARAFAC Decomposition (CPD) is studied. Firstly, the instantaneous hybrid signal decomposition is analyzed with an instantaneous hybrid model and a regular parallel decomposition method. Secondly, the signal separation problem applied in the instantaneous hybrid model is solved by using a regular parallel decomposition method from tensor decomposition theory. Finally, motivated by the idea of algorithm optimization by adding an operator, a new CPD method with proximal gradient constraint based on the classical CPD separation algorithms is proposed, which can be called as CPD-APG. Compared with the conventional JADE separation method and FastICA separation method, the separation error of the proposed algorithm is generally reduced by about 50% under the SNR range of-20dB~0dB. When the SNR is above 0dB, the separation error of the proposed algorithm is nearly close to zero.
KW - Candecomp/parafac decomposition
KW - blind signal Separation
KW - instantaneous hybrid model
UR - https://www.scopus.com/pages/publications/85149124424
U2 - 10.1109/ICISPC57208.2022.00024
DO - 10.1109/ICISPC57208.2022.00024
M3 - 会议稿件
AN - SCOPUS:85149124424
T3 - Proceedings - 2022 6th International Conference on Imaging, Signal Processing and Communications, ICISPC 2022
SP - 89
EP - 93
BT - Proceedings - 2022 6th International Conference on Imaging, Signal Processing and Communications, ICISPC 2022
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 6th International Conference on Imaging, Signal Processing and Communications, ICISPC 2022
Y2 - 22 July 2022 through 24 July 2022
ER -