Study On Synthetical Algorithm For Automatic Phase Correction Of NMR Spectra Based On Neural Network

Nuclear Magnetic Resonance(NMR), as a powerful tool to study the material structure, is widely used in the area of chemistry, physics, biomedicine etc. Because phase-distorted spectra can hardly be used in spectra analysis, phase correction of FT NMR spectra is one of the basic steps in NMR data processing. Though manual phase correction can often give satisfying results, automatic phase correction becomes more and more important due to the requirement of automatic batch processing. Ever since 1969, automatic phase correction has been attracting research interest.Nowadays, a number of automated phase correction algorithms have been proposed with the goal to replace the conventional, manual approach. However, due to the differences in the nature of algorithms and the characteristics of the spectra on which the algorithms based, different algorithms have different applicability to different spectra. After an extensive research of existing automatic phase correction algorithms, a synthetical algorithm for automatic phase correction of NMR spectra based on neural network, namely NNAPC, is proposed here. The proposed algorithm uses artificial neural network to choose an appropriate algorithm to calculate the phase angle of a given peak using the characteristics of the sepctra or the peak. NNAPC is proved by statistics to be more accurate and more stable than any existing algorithms referenced.There are four chapters in this dissertation. Chapter One includes an introduction of the theory of the phase correction and a review of previous work on automatic phase correction. Chapter Two describes the implementation of four automatic phase correction algorithms which will be used in our NNAPC algorithm. Chapter Three describes the proposed Neural Network based Automatic Phase Correction(NNAPC) algorithm. Chapter Four summarizes the work and gives some suggestion on future works.