Target Localization Based On Time Reversal Signal Processing

High-accuracy target localization has been more and more important in modern life.For example,in military application case,it has been widely used in target detection,localization and tracking,which can then help us to lock and destroy our enemies.In civil life,it is even more popular in vehicle navigation,urban search-rescue system and also an automatic guide in a shopping mall.However,the target to be localized is usually embedded in a very complex environment in which the received signal contains massive multipath components.As a result,how to get a high-accuracy position estimation of the target in the presence of multipath signals has been the interest of both academia and industry for several decades.In general,people tend to regard multipath as distortions in localization scheme and try to eliminate or mitigate the multipath as much as possible.However,this kind of methods degrades a lot under the complex terrains or in the urban areas in which the massive multipath components become a mess.Recently,time reversal(TR)has been proved to be able to compensate the multipath distortions by taking advantage of the reciprocity of the wireless propagation channels.Thus,the probing signal can focus on the source location and then be scattered back into the receiver.Evidently,focusing on the target increases the power ratio impinging on the target and then improves the later target localization performance.Inspired by such an idea,in this dissertation,we focus on how to develop a high-accuracy target localization scheme by using TR technique to adaptively exploit multipath signals in a rich-scattering environment.1.We explore a typical rich-scattering scenario and then propose a signal model in which the target is in low-angle and altitude.First,we study the impinging angle scope of both direct path and multipath signals.Then,we introduce the first-round received signal on the base of array signal processing and the aforementioned impinging angles of both direct path and multipath signals.According to the TR theory,we then apply TR processing on the first-round received signal,that is first transforming the first-round received signal into frequency domain by Fast Fourier Transform(FFT),then taking the conjugate and energy normalization to obtain the TR transmitting signal.Finally,we further consider the unique focusing property of the TR transmitting signal and then derive the TR received signal model.This model lays the foundation of how to use TR technique to solve the typical signal processing problems such as high-accuracy direction of arrival(DOA)estimation,target detection and so on.2.We propose a high-accuracy DOA estimation method for low-angle target under complex terrain.We first explore the parameters related with the signal model such as impinging angle of the direct path,impinging angle of specular reflection signals,impinging angle of diffuse reflection signals and the synthetic reflection coefficients including reflection coefficient,divergence factor and surface roughness.Then,based on the meter wavelength signal,we propose a new low-angle target DOA estimation method by using the TR technique.Specifically,Spatial smoothing is taken to resolve the coherence between the direct path and the multipath signals.TR technique takes advantage of multipath echoes recorded by a sensor array and adaptively adjusts TR probing waveforms to increase the signal-tonoise ratio(SNR)in a rich-scattering environment.Moreover,the coherent signal-subspace method(CSM)compresses the energy of a signal into a predefined subspace to exploit the full time-bandwidth product of signal sources and cope with coherent wavefronts.Finally,Cramer-Rao bounds(CRB)analysis and numerical simulations commendably validate the superiority of the proposed method in dealing with multipath signal and improving localization accuracy.3.Given that the TR technique takes wideband signal as the probing signal in most cases,we develop a DOA estimation method for low-angle target in frequency domain.First,we analyze the ambiguity between the DOA and frequency information because it will degrade the accuracy of steering vector based DOA estimation method such as digital beamforming,multiple signal classification(MUSIC)method.To solve it,we introduce a virtual delay factor which is independent of the DOA information to eliminate the ambiguity.Next,we analyze the independence between the noise subspace and the sampling frequency bin in our TR wideband signal model.By using noise subspace,we can greatly reduce the computational load needed to compute the covariance matrix and corresponding inverse matrix.Moreover,perfect wireless channel is usually assumed in traditional TR based methods,which is not always true in practice.As a result,we introduce a random perturbation factor in the signal model to represent the channel variety.And in beam domain,a constant beamforming matrix is applied to alleviate the impact of the perturbation on the target estimation accuracy,which yields a high-accuracy target localization algorithm in beam domain.Finally,the superiority of our method in accuracy,complexity,solving the frequency and DOA ambiguity and robustness in non-stable environment for low angle targets are validated by numerical simulations.4.We extend the TR technique into civil applications and combine it with the future 5G massive MIMO especially in urban area which is full of multipath signals.Different from the localization model for low-angle targets,civil targets such as vehicles and cellphones are always able to communicate with surrounding base stations,which can be then used for localization.Meanwhile,implementing TR processing physically takes a re-transmitting and receiving signal process,which greatly hinders the applications of TR based methods.In this case,we first set up the signal model for the urban targets on the base of massive MIMO scheme.Then,we prove that TR technique is equivalent to the auto-correlation function(ACF)of the received signal around the receiver in spatial domain.Thus,we use the ACF of the received signal rather than doing TR processing physically and develop a high-accuracy target moving direction estimation method by further using the distribution of 5G base stations.Finally,we explore the relationship between the energy distribution of the autocorrelation of the received signal and the location information of the base stations.Given the prior information about the location of the surrounded base stations,this dissertation proposes a high-accuracy target localization method method which can achieve centimeter accuracy at the best and at least sub-meter accuracy in the presence of severe multipath reflections.In the end of this paper,we give the summarization and make some plans for the future.