Research On Random Matrix Based Method For Target Detection In MIMO Radar Under Correlated Noise

Compared with traditional radar systems,multiple-input multiple-output(MIMO)radar can use multiple antennas for transmit diversity and receive echo signals from multiple antennas,thus the performance of target detection can be significantly improved.At present,MIMO radar target detection methods mainly include generalized likelihood ratio test(GLRT),constant false-alarm rate detection and matched filter detection,etc.Despite the fact that these methods have improved the detection performance in different degrees,they still have the following main drawbacks:(1)the number of snapshots is assumed to be far more than the number of antennas.However,it is no longer applicable when the number of antennas is large and similar to the number of snapshots;(2)the noise variance is required known or estimated in advance,which will lead to large errors when the number of snapshots is insufficient;(3)noise model is usually simplified as ideal white noise.However,in actual environment,the mutual coupling among antennas and oversampling may lead to correlated noise.If the detection method and the corresponding decision threshold for white noise are still used,the detection performance will be seriously degraded.To solve the above problems,taking random matrix theory(RMT)and canonical correlation technique(CCT)as mathematical tools,target detection methods for MIMO radar in correlated noise are investigated in the thesis.Aiming at transmit diversity MIMO radar system,and considering that the numbers of receiving array elements and snapshots grow at the same rate,random matrix-based target detection method in white noise is first analyzed.On this basis,random matrix-based target detection methods in the backgrounds of spatial correlated noise and spatial-temporal correlated noise are proposed,which are suitable for large array system.The work is supported by the National Natural Science Foundation of China “Robust target detection and estimation for MIMO radar based on large dimensional random matrix theory”(No.61371158).The creative works of the thesis are as follows:A CCT-based target detection method for MIMO radar in spatial correlated noise is proposed.In this method,two separate sub-arrays are used as MIMO radar receiving arrays.Assuming that the elementary noise in each sub-array has spatial correlation,the received signal model with two sub-arrays is generated.Furthermore,the detection statistics is constructed based on GLRT criterion and canonical correlation factors between two sub-arrays,and the expression of decision threshold is deduced with the second distribution of Tracy-Widom(TW)in RMT.The simulation results show that the detection performance of this method is better than that of the traditional conditional number(CN)algorithm in the presence of spatial correlated noise and large arrays.Further,RMT-based target detection methods for MIMO radar in spatial-temporal correlated noise are put forward.Firstly,the spatial-temporal correlated noise model is generated using correlation coefficient matrix.Then,the detection statistic is constructed using the normalized maximum eigenvalue(ME)of the sample covariance matrix of the received signals.Then the TW distribution in RMT is used for target detection method in MIMO radar in spatial-temporal correlated noise—TW-ME algorithm.The second distribution of TW is exploited as the asymptotic distribution of the maximum eigenvalue which is simulated by the normalized covariance matrix of the received signal samples.The relationship between decision threshold and false alarm rate is deduced under the background of space-time correlated noise.Furthermore,the TW-ME algorithm is improved by combining the semi-circular law with the M-P law.Also,Stieltjes transform is used to modify the parameters in the decision threshold,thus a new threshold is obtained.The simulation results show that the improved TW-ME algorithm has better detection performance.