Construction of Roc Curves of Radio Location Signals With Detection Threshold According to the Hurst Index
Year: 2025
Type: article
Abstract: The ROC curve (receiver operating characteristic) is a modern analytical tool used in classification tasks, displaying the true-positive and false-positive rates. It is considered the best diagnostic test method because it shows the best cut-off value for diagnostic performance. In the paper, for the first time, an ROC-curve is constructed, which determines the characteristics of a signal receiver with a fractally dependent coefficient of detection of radar signals using a continuous wavelet decomposition of signals from the time domain in the frequency domain and with the determination of fractal features using the generalized Hurst coefficient. The ROC curve constructed by the mentioned method showed the maximum proximity to the upper left corner of the graph, indicating a high true-positive rate and accurate determination of positive results. The use of the classic Hurst coefficient as a fractal feature revealed proximity to the diagonal of the graph, confirming the low quality of detection. It is shown that for 2D images of wavelet spectra, the Minkowski dimension obtained by the box counting method can be used as a fractal feature, exhibiting the maxima of the fractal dimension and the noise power. This made it possible to identify the image of the spectra. The threshold for such identification was determined based on a comparison of the autocoherence of the series of wavelet coefficients from which the specified images are generated. Keywords: RОС-analysis of signals, fractal dimension, Hurst coefficient, Minkowski fractal dimension, wavelet-decomposition, radar signal.
Source: Kibernetyka ta Systemnyi Analiz
Authors: Yu. Taranenko, O. Oliinyk, Y.V. Khomyak
Language: English
Subfield: Aerospace Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W7082358258