Nonparametric Method for Signal Detection Using SG Statistics

The paper proposes a new nonparametric method for signal detection based on Savit and Green (SG) statistics and signal observation processing using the surrogate data technology, namely Attractor Trajectory Surrogates (ATS) algorithm. The proposed detection method does not depend on a priori information about the interference probability distribution density and does not consider the model of the observed signal. Using the ATS algorithm allows us to reduce the noise in signal observation. The noise observation estimate has been obtained from the observation itself using the singular spectral analysis for the first time. A comparative analysis of the proposed nonparametric and classical signal detection methods using an energy receiver based on χ²-statistics has been performed. The simulation results showed an increase in the probability of signal detection using the SG statistics-based method for different signal-to-noise ratios compared to using an energy detector. It is also demonstrated that the proposed nonparametric detection method does not depend on the density of the interference probability distribution, using an example of interference with a logistic distribution. Recommendations are given for selecting the parameters of the proposed nonparametric detection method.