Time Series Complexity Estimation Method Based on Image Microstructure Frequency Analysis

Abstract

A time series complexity calculation method based on image microstructure frequency analysis: construct a recursive matrix of the signal; draw a grayscale image, draw a grayscale image according to the recursive matrix of the recursive state at the i-th moment and the j-th moment; for the grayscale image , use Gaussian kernel functions of different scales to filter, obtain different Gaussian grayscale images, form a Gaussian pyramid, and perform convolution operations on Gaussian kernel functions of different scales and grayscale images to obtain the scale space of the image; initially determine the features in the scale space Point position and scale; use the Taylor function quadratic expansion of the scale space function to perform least square fitting, use the extreme value of the fitted surface to remove unstable feature points; cluster the remaining feature points; continuously change through calculation The values ​​of influence degree and validity degree; measure the information of clustering results; use approximate entropy and permutation entropy to calculate the complexity of different signals, and compare and analyze the results with the results of microstructure recursive entropy. The invention provides useful reference for the design of new entropy method.

Description

technical field [0001] The invention relates to a method for measuring and calculating the complexity of time series. In particular, it relates to a time series complexity measurement method based on image microstructure frequency analysis. Background technique [0002] Complexity science, which emerged in the 1980s, is a new stage in the development of system science and one of the frontiers of contemporary scientific development. The development of complexity science has not only triggered changes in the natural sciences, but has also increasingly penetrated into the fields of philosophy, humanities and social sciences. With the popularity of big data, data-driven methods have become an important means of studying complex systems, and time series analysis has become a powerful tool for exploring and interpreting the inner operating mechanism of complex systems. Related achievements have been widely used in physics, economics, Meteorology, linguistics and information scie...

AI Technical Summary

Problems solved by technology

The study found that the existing methods are mainly limited to design entropy methods in the time domain and frequency domain, so the complexity analysis is somewhat one-sided

Images