O.F. Antonov, Y.A. Sergeev, S.G. Kovalenko

Statement of the problem. The detection of a person hidden behind a radio-transparent barrier is of interest for various applications: search and rescue of victims, security systems, vehicle inspection, anti-terrorist activities, and others. To solve the problem of detecting a person hidden behind a radio-transparent barrier by the method of bioradar, it is necessary to develop a detection model based on informative features that take into account both the characteristic movements of a person and the physiological processes of the human body, inherent even to a motionless person.

Target. To consider the process of creating a model for detecting a person hidden behind a barrier based on the signs of his vital activity identified with the help of a bioradar, based on the generated feature space, as well as optimizing its parameters.

Results. The process of forming a feature space for the tasks of detecting a person based on the accumulated base of bioradar signals is presented. The formation of informative features was carried out taking into account the difference in the level of amplitude values in the respiratory frequency range for useful and background signals, as well as the difference in the mathematical expectation and dispersion of amplitude values in the human respiratory frequency range and other frequency ranges. A detection model based on informative features for the problem of detecting a person with a sufficient confidence probability is presented in the form of a formula for calculating the value of the feature and the threshold value of the feature for making a decision according to the criterion of an ideal observer. Detection characteristics were obtained using one feature and a vector of five best features. The detection accuracy for one feature exceeds the value of 0.91, the detection accuracy for five features exceeds the value of 0.95. Data validation was carried out using the leave-one-out cross-validation method. A technique is proposed for optimizing eight parameters of the obtained model, which are taken into account both when processing received signals and when extracting informative features. The difficulty in finding optimal values for frequency bands was overcome by combining well-known optimization methods based on Bayesian statistics with calculating the classification accuracy for each frequency band. As a result of applying the optimization technique, the values of the parameters of the detection model were found, providing the indicated classification accuracy.

Practical significance. The approaches used in the study to the formation of a feature space for detecting a person hidden behind a radio-transparent barrier, creating a detection model based on informative features and optimizing the parameters of the resulting model can be applied to create detection subsystems in the development of bioradar in order to solve the problem of detecting a person hidden behind a barrier in various areas.

Antonov O.F., Sergeev Y.A. Kovalenko S.G. A model for detecting a person hidden behind a barrier using a bioradar. Radiotekhnika. 2023. V. 87. № 2. P. 26−33. DOI: https://doi.org/10.18127/j00338486-202302-04 (In Russian)

1. Yang D, Zhu Z, Zhang J, Liang B. The Overview of Human Localization and Vital Sign Signal Measurement Using Handheld IR-UWB Through-Wall Radar. Sensors. 2021. V. 21(2). P. 402.
2. Modelirovanie i obrabotka radiolokacionnyh signalov v Matlab: Ucheb. posobie. Pod red. K.Yu. Gavrilova. M.: Radiotekhnika. 2020.
   264 s. (In Russian).
3. Anishchenko L., Ruckova E. Formirovanie prostranstva priznakov pri raspoznavanii stadij son-bodrstvovanie laboratornogo zhivotnogo pri pomoshchi bioradiolokatora. Biomedicinskaya radioelektronika. 2017. № 8. S. 48–54 (In Russian).
4. Shalev-Shvarc SH., Ben-David Sh. Idei mashinnogo obucheniya: ot teorii k algoritmam: Per. s angl. A.A. Slinkina. M.: DMK Press. 2019. 436 s. (In Russian).
5. Breiman L. Random Forests. Machine Learning Journal. 2001. V. 45. № 1. P. 5–32.