A review of NQR signal processing and analysis techniques

Nuclear quadrupole resonance is a solid-state spectroscopy technique used for detecting compounds containing quadrupolar nuclei. Its advantages, such as high specificity and non-invasive scanning, have enabled its application in pharmaceutical analysis and prohibited substances detection. However, it has the disadvantages of very low signal-to-noise ratio and temperature dependence of the response signal, and multiple signal processing techniques have been developed to compensate them. The purpose of this article is to provide a comprehensive review of the signal processing and analysis techniques developed for improving detection in nuclear quadrupole resonance spectroscopy, in order to serve as a starting point for new researchers. An introduction in this field is given and the specific signal processing challenges are presented. A new classification of existing solutions is proposed and a comparative analysis is performed in different noise and temperature conditions on multiple statistical and machine learning algorithms. The classification solutions are shown to outperform the statistical ones and the use of machine learning techniques is proposed for future improvement. This study is also instructive for research regarding the selection of a signal processing and analysis algorithm for nuclear quadrupole resonance detection and can be used as a reference for future works.