Research And Application Of Wideband Signal Blind Source Separation Algorithm

Wideband blind source separation refers to the process of recovering the source signals only according to the characteristics of the observed signals and all parameters of the source signals and transmission channel are unknown when the source signals are wideband signals.The existing blind separation algorithm for wideband signals is not perfect,so the problem of wideband blind separation can be solved by using wideband blind beamforming.Wideband blind beamforming means that the desired signal extraction can be realized only through its own algorithm when the signal direction and array popularity are unknown.Therefore,this paper mainly studies the wideband blind separation algorithm based on blind beamforming and its application.The main research contents are as follows:1.To solve the problem of poor performance of the traditional blind separation algorithm for wideband array signal separation,a wideband blind separation algorithm based on TDL(Tapped Delay Line)blind beamforming is proposed.Firstly,the adaptive noise reduction,single source point detection and delay estimation are performed on the observed signals,and the received signal is compensated for the pre-delay.Finally,the wideband blind beamforming algorithm is used to extract the different expected signals to realize wideband blind separation.Then a low complexity wideband blind separation algorithm based on no TDL blind beamforming is introduced.In this algorithm,the wideband signal of one direction is equivalent to the narrowband signals of several adjacent directions,and finally the wideband signal is extracted by narrowband beamformer,which greatly reduces the complexity of the algorithm.The simulation results show that the two algorithms can achieve a good separation effect,and there is no fuzzy problem of amplitude and ordering.On the other hand,due to the complexity of FH(Frequency hopping)signal which is instantaneous narrow band and macroscopic wide band,the existing wideband blind separation algorithm has poor adaptability.In this paper,the effectiveness of the two algorithms for FH signals with large number of overlapping time-frequency points is verified by experiments.2.Aiming at the problem that errors in delay estimation lead to errors in pre-delay compensation and result in the distortion of output waveform,a wideband blind separation algorithm based on constrained least squares blind beamforming is proposed.Firstly,a sectionalized adaptive noise reduction algorithm is proposed to estimate the time delay effectively when the power of the input signals is different.Then based on the conventional least square beamforming algorithm,a wideband blind separation algorithm based on improved constrained least square blind beamforming is proposed.First the single source detection and delay estimation algorithm are adapted to obtain the delay,then the spatial response variation(SRV)constraint and the modular constraint are introduced to extract the expected signal,then the wideband blind separation can be realized.The simulation results show that the algorithm still has good separation effect when the power of the incident signal is different,the added constraint conditions enhance the robustness of the algorithm greatly,and the problem of inconsistent frequency response is solved well.3.In view of the traditional FH network station sorting method is difficult to implement under a complex electromagnetic environment,the wideband blind source separation algorithm is applied to the FH network station sorting in this chapter.First the wideband blind source separation algorithm is used to obtain source signals preliminary,and then the time-frequency processing is performed on the separated FH signal to make it separated completely,which realize network sorting automatically.Finally the hop instants,the hop duration and the frequencies are estimated which are obtained by a single FH signals,and then evaluate the performance of the network station sorting by observing the parameter estimation accuracy.Compared with the traditional sorting method,this method has nothing to do with the modulation mode of FH signal and it doesn't matter whether there's a collision or not,and the automatic sorting of FH network station can be realized by blind source separation with low complexity and high flexibility.