Study On The Parameter Estimation Methods Of Weak Nuclear Quadrupole Resonance Signals

Nuclear quadrupole resonance (NQR) is a solid-state radio frequency (RF)spectroscopic technique, allowing the detection of various high explosives and illicitmaterials. Unfortunately, the practical use of NQR is restricted by the low signal tonoise ratio (SNR) and the radio frequency interference (RFI). The interferencemitigation, parameter estimation and detection of weak NQR signals are the core andthe key of the NQR technique. From the perspective of signal processing, some topicsrelated to NQR, such as the basic concept, the signal model, the signal excitationmethods, the interference mitigation, and parameter estimation and detection of weakNQR signals, are analyzed and elaborated comprehensively in this dissertation. Theinterference mitigation, detection and parameter estimation of weak NQR signals areresearched deeply. The specific work is summarized as follows:1For the problem of interference mitigation and weak NQR signal detection, amethod of interference mitigation based on singular value decomposition (SVD) of theHankel matrix is proposed. In the proposed algorithm, a Hankel matrix is formed byoriginal data at first, and its SVD is implemented, then the original signal can bedecomposed into several component signals according to the singular values’ order.Finally, in accordance with the spectrum estimation of each component signal, theinterference and noise are removed on the basis of the prior information of NQR signals.It suppresses the interference effectively, and improves the SNR. After filtering theinterference, a nonlinear least square (NLS) detector based on MUSIC spectralestimation is presented, with the generalized likelihood ratio test (GLRT) as a detectionvariable. The proposed algoriyhm greatly reduces the computation load, simultaneously,guarantees a superior resolution of the frequency, so the motivation of the availabledetection for weak NQR signals is achieved.2Be aimed at the parameters estimation problem of the free induction decay(FID) signal contained interference and thermal noise, the linear prediction estimatorbased on the total least squares (TLS) is presented to estimate the parameters of the FIDsignal. The perturbations of the linear prediction equation on the two sides areconsidered in the proposed method, and then solved in the total least squares sense, withthe consideration of the interference and noise terms, the drawback of the conventionalleast squares method which only considers one side disturbance and leads to a relativelylarge variance, is overcomed. With the combination of the damp character of the FID signal and the criterion of the maximum likelihood estimation, the estimation of theparameters of the weak FID signal is effectively accomplished.3In allusion to the methods of parameters estimation of weak NQR signalsmasked by strong interference signals, residual signal iterative analysis algorithm(RSIAA) based on the improved fast maximum likelihood estimation (IFMLE) isproposed to estimate the parameters of the NQR signals. The proposed method turns themultidimensional search problem into multiple one-dimensional searches, and combinesthe interpolator and semi-Newton method. It solves the mask problem of theinterference to the NQR signal effectively, which simultaneously reduces thecomputational complexity and improves the estimated precision. The effectiveness ofthe proposed algorithm is demonstrated by the processing results of both simulated dataand experimental data.4For the time-variation characteristic of interference signals in the explosivedetection using the NQR technique, and according to the real world signal model ofeach signal sample block including the fixed NQR signal, time-variation interferencesignals and noise, an improved multi-block Hankel total least squares stack (HTLSstack)method is presented. Firstly, the number of common poles is determined using theHTLS scheme; secondly, a new selection criterion is constructed to choose the commonpoles; finally, the prior knowledge is employed to realize the estimation of NQR signalseffectively. This method makes full use of the information of original data, and in thecase of low signal-to-noise ratio, it avoids the performance loss of multi-block HTLSmethod in the first SVD, and improves the precision of parameter estimation availably.5To the problem of the multi-parameter estimation of a spin echo train (SET)signal in the presence of additive white Gaussian noise, a two-dimensional (2-D)parameter estimation method in a separable manner is proposed. By utilizing therank-one property of the2-D data matrix, and according to an iterative weighted leastsquares (WLS) method, the damping factor and frequency are estimated separately fromthe principal left and right singular vectors. The complex amplitude of SET is thenobtained on the basis of standard least squares. The parameter estimation performanceof this method achieves Cramer-Rao lower bound (CRLB) at a relatively large SNR ordata size conditions, and has a low computational complexity.6In allusion to multi-parameter estimation of multi-spectrum SET signal in thepresence of additive white Gaussian noise, a two-dimensional (2-D) parameterestimation method based on eigenvector is proposed. In this method, by using the SETsignal, a2-D data matrix is constructed, and according to the different manners to form the Hankel block matrix pencil (MP), and the eigenvector is solved by utilizing theshift-invariance structure of the signal subspace, and further the damping factor as wellas the frequency are estimated with the structure rule of the eigenvector. The methodhas the capability of auto association, and under a relatively large SNR and availablefrequency resolution conditions, it can estimate the parameters effectively. Simulationresults show the effectiveness of the proposed algorithm.