Study On Single Channel Blind Source Separation Algorithm Based On Dual-Tree Complex Wavelet Transform

Automatic Identification System(AIS)is a broadcast automatic reporting system that guarantees the safety of maritime transportation and improves the efficiency of maritime communication.Shipborne AIS uses ships to carry AIS devices,which is the primary form of AIS.Afterwards,spaceborne AIS is proposed in order to expand the field of the AIS's view and improve the effectiveness of information transmission.Due to the short development of spaceborne AIS,there are still many key technical challenges to overcome,such as message collision problems,electromagnetic wave loss and delay in long distance transmission etc.This thesis mainly focuses on the single channel blind source separation(SCBSS)problem in spaceborne AIS.The details are as follows:(1)A SCBSS algorithm based on dual-tree complex wavelet transform(DTCWT)is proposed.Firstly,DTCWT is used to decompose the received signal,and then some signal components are filtered out from the decomposed signals by PCA,whose number is one less than that of the source signals.After that,the filtered signals constitute the virtual multi-channel mixed signals together with the received signal.Finally,some of the source signals are estimated by Fast-ICA.This algorithm greatly reduces the frequency aliasing problem in traditional wavelet decomposition.Simulation results show the performance of the proposed algorithm is significantly improved compared with the SCBSS algorithm based on wavelet transformation.(2)A SCBSS algorithm based on lifting Q-shift DTCWT is proposed.Q-shift DTCWT can solve the problems of the different frequency response between the two trees caused by poor structural symmetry in general DTCWT.While the lifting algorithm converts the arithmetic of Q-shift DTCWT into time domain,which simplifies computational complexity and saves internal memory and time.The simulation results show that the algorithm can shorten the running time of the program and improve the real-time performance without affecting the separation effect of the SCBSS algorithm based on DTCWT.(3)A SCBSS based on DTCWT and EEMD is proposed.The algorithm uses DTCWT to decompose the received signal.And then part of decomposed signals are processed by EEMD to obtain some groups of IMFs.After that,select some signals from IMFs to constitute the virtual multi-channel mixed signals together with the received signal.And these virtual multi-channel mixed signals are separated by Fast-ICA to obtain the estimated source signals.The simulation results show that compared with the SCBSS algorithms based on EEMD and DTCWT respectively,the proposed algorithm improves the separation performance at high SNR,at the same time,it has the anti-noise performance of SCBSS algorithm based on DTCWT in low SNR.