Surrogate Data And Its Application

For random signal processing, the test of non-linearity as well as the stationarity test is often an important and basic pre-processing step. This paper revisits the surrogate method, which is a method using some constrained noise as a reference for statistical testing. In detail, this paper introduces several methods for creating surrogates and the use of surrogates in the tests of non-linearity and stationarity.In addition, this paper explains our design and implementation of a system for random signal analysis based on MATLAB. This system gives an interface for basic statistical test, the construction of surrogate using assigned method and providing the non-linearity features and the characteristics in time-frequency domain of both original data and surrogate data.In the test for non-linearity, the non-linearity features of original data and surrogate data are computed by the system. Then the difference between those features can be used for non-linearity test.In the test for stationarity, the original data are stationarized by creating surrogate data. Then the differences between the characteristics in time-frequency domain of original data and surrogate data can be used for stationarity test.This system gives a friendly interface and is convenient to use. The experiment indicates this system can perform the test of non-linearity and the stationarity test well.