No abstract

The received signal strength difference (RSSD) localization is a kind of method to locate emission sources by measuring the differences of received signal strength level between the monitoring stations and is essentially the truth value ratios of measured signal strength. In the existing literatures, only the rule of RSSD localization circle of two monitoring stations and the geometric relation of RSSD localization circle of five monitoring stations were analyzed, but the number and the station layout of the minimum RSSD localization network have not been investigated. In the present work, first, based on the existing RSSD localization equation, the constants of the commonly used wave propagation models are provided. Then, the minimum RSSD localization network is proved through algebraic analysis, which is that four monitoring stations not distributed on a straight line can locate the signal source at one point. The relationship between the localization accuracy and the signal strength error of the RSSD location network with different scales is studied further and formulated as a nonlinear programming optimization problem. It is found that the localization stability of the network with four stations is poor, and there is a serious localization deviation outlier phenomenon. Therefore, the network with four stations is not available for radio monitoring networks with a signal strength error of ± 5 to  ± 10 dB. The RSSD network with five stations is basically the minimum available size, and the RSSD network with nine stations can approach the localization accuracy of the angle of arrival (AOA) network with three stations.

This paper presents an improved genetic algorithm (GA) based feature selection method for imbalanced data classification, which is then applied to radio signal recognition of ground-air communication. The proposed method improves the fitness function while SVM is selected as the classifier due to its good classification performance. This method is firstly evaluated using several benchmark datasets and experimental results show that the proposed method outperforms the original GA-based feature selection method now that it not only reduces the feature dimension effectively, but also improves the precision of the minor class. Finally, the proposed method is applied to a real world application in radio signal recognition of ground-air communication, which again shows comparatively better performance.

No abstract