Research On Technique Of Array Signal Processing Based On Time-Frequency-Spatial Information

The array signal processing technique can estimate the direction of arrival(DOA)and extract the desired signal even in a complex electromagnetic environment.It has many advantages,such as high super-resolution,high gain,and simultaneous multi-target processing.However,in practical application,there always exists sensor position error,gain/phase error,limited snapshots,etc.They will degrade the performance of DOA estimation and spatial filtering severely,and it is difficult to achieve significant performance improvement only with the spatial information.Therefore,in this dissertation,we introduce the time-frequency(TF)analysis technique into array signal processing and focus on the wideband array self-calibration and DOA estimation under sensor position error,narrowband adaptive beamforming under random error,wideband adaptive beamforming under DOA error and sensor position error,and low-complexity wideband beamforming,to improve the performance of DOA estimation and beamforming in practical application.The main works of this dissertation are summarized as follows:1.In the presence of sensor position error,existing self-calibration methods have high computational complexity and cannot work well for wideband signals,whose performance will degrade severely with the error increasing.Hence a novel wideband self-calibration and DOA estimation method is proposed,which is based on the short-time Fourier transform(STFT).We firstly transform the received signal into TF domain to separate the sources better,and cluster the TF points based on spatial features.Then,the single-source TF points of source signals are selected,which are utilized to calculate the propagation delays.Due to the simple linear relationship between the propagation delays and sensor positions,the DOA and true positions are finally estimated by a two-step iteration without high-dimensional search or global search,guaranteeing faster convergence rate.Simulation results demonstrate that the proposed method can achieve accurate DOA and position estimation even under large sensor position errors with low computational complexity.2.The conventional robust adaptive narrowband beamforming algorithms suffer severer performance degradation with the increase of input signal to noise ratio(SNR),and the existing covariance matrix reconstruction-based beamformers have poor robustness against random array calibration error.Therefore,a novel adaptive narrowband beamforming method is proposed,which is based on the quadratic TF transform.Firstly,it obtain the TF information of the narrowband source signals using the quadratic TF transform,which has higher resolution.Then,it reconstructs the interference-plus-noise covariance matrix and estimates the steering vector of the SOI using the spatial time-frequency distribution(STFD)matrices at the single-source auto-source points of the desired signal and interferences,respectively.The final beamforming weights are calculated via minimum variance distortionless response(MVDR)method.In the whole procedure,the imprecise prior information about the array manifold is not utilized,hence the random error cannot degrade the output performance.Simulation results demonstrate that the proposed method has strong robustness against random error,and achieves high output signal-to-interference-plus-noise ratio(SINR)close to the optimal one.3.In the case of direction error,sensor position error,and limited snapshots,the existing adaptive wideband beamforming methods cannot achieve good enough output performance.Hence a novel adaptive wideband beamforming method is proposed,which is based on the quadratic TF transform.Firstly,the propagation delays under direction error and sensor position error are estimated using the TF spatial features of sources,with which the steering vector of the desired signal can be easily obtained.Then,the covariance matrices of interferences at different frequencies are calculated via Capon spectral spectrum,and the wideband interference-plus-noise covariance matrix is reconstructed by summing these covariance matrices.Finally,the beamforming weights are obtained by the wideband beamformer without pre-steering delays.Besides,we propose another robust wideband beamforming method based on STFT,which can reconstruct the interference-plus-noise covariance matrix using the linear TF distributions directly,and can also estimate the steering vector of desired signal precisely.Simulation results demonstrate that the proposed methods are robust against the direction error and sensor position error,and it can achieve high SINR close to the optimal one with small number of snapshots.4.The tapped delay-lines(TDLs)and sensor delay-lines(SDLs)in wideband beamformer improve the system complexity and computational cost.To reduce the complexity,a novel beamformer without TDLs or SDLs is proposed.It firstly treats one wideband signal from a direction as multiple narrowband signals from an angular sector with the same equivalent frequency.Then it makes the desired signal come from by compensating its propagation delays,so that the desired signal have the same steering vector at different frequencies.Finally,the wideband spatial filtering is realized by narrowband beamformer.To improve its robustness,the wideband beamformer is further combined with the TF spatial method.Simulation results demonstrate that eliminating the TDLs and SDLs leads to a simpler system and higher efficiency,and the proposed method can achieve satisfactory output performance and fast convergence rate even in the case of direction error and sensor position error.