Terahertz Spectral Identification Method Combining Radial Basis Function and Kernel Principal Component Analysis

Abstract

The invention discloses a terahertz spectrum identification method combined with radial basis function and kernel principal component analysis, which is used to effectively reduce the dimension of terahertz spectra and can solve the classification problem of terahertz spectra when they are linearly inseparable. The test substance is subjected to sample preparation, and then the sample of the test substance after sample preparation is put into the terahertz time-domain spectroscopy system to obtain the terahertz spectrum of the test sample, and the moving average filter method is used to remove the terahertz spectrum of the test substance. Rough processing, then use radial basis function mapping, and finally use nuclear principal component analysis to perform dimension reduction processing to obtain features, and input the features to a trained support vector machine for classification to obtain material recognition results. The invention proposes a method for extracting terahertz spectral features by combining radial basis function and kernel principal component analysis, and the feature clustering effect extracted by the method is good, and the classification accuracy is high.

Description

technical field [0001] The invention relates to the field of substance identification, and more particularly, to a terahertz spectrum identification method combining radial basis function and nuclear principal component analysis. Background technique [0002] The terahertz spectrum of matter is mainly determined by the vibrational and rotational modes of matter intermolecular and intramolecular. Substances sometimes show obvious absorption characteristics in a certain terahertz frequency band, while other frequency bands do not have such obvious absorption characteristics. Therefore, there is a certain correlation between the various data variables of the terahertz spectrum of matter, and there is a lot of redundant information. Existing widely used principal component analysis technology mainly uses the idea of ​​dimensionality reduction, in the case of minimizing the loss of information, the original data is mathematically transformed to obtain a few features with statist...

AI Technical Summary

Problems solved by technology

[0003] In order to overcome the defect that the extracted features described in the prior art are not sufficiently distinguishable, the present invention provides a terahertz spectrum identification method combined with radial basis function and kernel principal component analysis

Images