【作者】 郭峰；

【导师】 马洪；

【作者基本信息】 华中科技大学 ， 无线电物理， 2012， 硕士

【摘要】 本论文的主要工作是研究非稳态核磁共振实验中的信号处理方法，其中包括通过自适应算法抵消射频干扰和利用谱幅度估计方法降低本底噪声，这些都可以提高获得的核磁共振信号的信号——噪声干扰比（SINR），最后利用足够高SINR的核磁共振信号反演出磁场的时域变化曲线。在本论文设计的自适应干扰抵消算法中，引入了三阶累积量，其既能够抑制高斯噪声和其他对称分布的噪声或者干扰，又能够提高算法的精度。设计出了一套基于三阶累积量的LMS(Least Mean Square)算法，并采用最优变步长作为步长因子，能够极大的提高算法的收敛速度。并通过仿真验证了算法的可行性，表明了算法具有很高的收敛精度，较快的收敛速度，通过该算法可以很大程度上提高SINR。自适应干扰抵消之后，利用之前获取的本底噪声来进行减谱降噪，在一定程度上消除本底噪声的影响，进一步提高SINR。在获取了足够高SINR的NMR(NuclearMagnetic Resonance)信号后，通过短时傅里叶变换对NMR信号进行时频分析，获取NMR信号各个时刻对应的频点，然后利用核磁共振条件，来反演磁场的整个时域变化曲线。仿真的结果表明，本论文的算法能够抵消掉较强的射频干扰，能够很好的反演出磁场的时域变化曲线。更多还原

【Abstract】 The main work of this paper is to research the signal processing methods ofnonstationary nuclear magnetic resonance experiment, including using the adaptivealgorithm to eliminate RF interference and using spectral subtraction method to reduce thebackground noise, which can improve the NMR signal--to interference and noise ratio(SINR), finally using the high enough SINR nuclear magnetic resonance signal inversemagnetic domain change curve.In this paper the design of adaptive interference cancellation algorithm, introduced thethree order cumulant, which can suppress Gauss noise and other noise or interference withsymmetric distribution, and can improve the accuracy of the algorithm. This paperdesigned a LMS(Least Mean Square) algorithm based on the three order cumulant, andused the optimal variable step length as step length factor, which can greatly improvingthe convergence speed of the algorithm. Through the simulation, the feasibility of themethod is proved; showed that the algorithm has high convergence precision, fastconvergence speed, the algorithm can greatly improving the SINR.After adaptive interference cancellation, this paper used the previously acquiredbackground noise to reduce noise by spectral subtraction, eliminated background noise toa certain extent, and further improved the SINR. After acquired the NMR(NuclearMagnetic Resonance) signal with sufficient high SINR, this paper used the short-timeFourier transform as time-frequency analysis method to analysis NMR signal, thenobtained the NMR signal corresponding frequency of each time point, and then usingnuclear magnetic resonance condition, to inverse the whole time domain curve of themagnetic field. The simulation results show that, the algorithm can counteract the strongradio frequency interference and can get a very good magnetic curve.更多还原