Novel grey correlation classifier design method

Abstract

The invention provides a novel grey correlation classifier design method comprising the following steps: a grey correlation classifier is designed by use of a gray correlation algorithm; and the initial images of the sequences of the behavior of a system are normalized and equalized, different identification coefficients are given to different signals, and therefore, the ability to restrain the identification coefficients is improved by use of an adaptive identification coefficient.

Description

technical field [0001] The invention relates to the fields of electronic countermeasures, signal recognition, and classifier design, and more specifically, the invention relates to a design method of a novel gray relational classifier. Background technique [0002] Today, with the continuous improvement of anti-reconnaissance and anti-interference technology, the complexity of communication systems and the continuous increase of noise, the individual signal differences are gradually reduced, and the traditional template comparison method has been difficult to complete the task of individual identification of radiation sources. How to realize the identification and classification of radiation source signals in the environment of low and unstable signal-to-noise ratio, the design and selection of classifiers become very important. The main function of classifier design is to make corresponding judgments according to the extracted signal features, so as to realize the classific...

AI Technical Summary

Problems solved by technology

For the traditional Deng's gray relational classifier, the anti-noise ability is relatively poor, and it is difficult to achieve the recognition effect under the condition of low signal-to-noise ratio; the T-type correlation degree uses the speed ratio to reflect the development trend of the two sequences, and the slope correlation degree is The speed difference is used to reflect the similarity of the development trend of the two sequences or the shape of the curve. However, the dimensionless process of the original data actually changes the proportion of the curve, so the slope correlation degree does not meet the standardization; the B-type correlation degree comprehensively utilizes Displacement difference, velocity difference, acceleration difference to reflect the similarity and similarity of the two sequence curves, focusing on the overall analysis

Moreover, the recognition rate of the above models for signals loaded with signal-to-noise ratio is not high

Images