Integrated Recognition on Overlapping Spectrum Signals in UAV-Assisted Radio Monitoring Systems

Due to the convergence of new wireless technologies progress and integration with the existing radio spectrum, overlapping spectrum signals (OSS) appearing simultaneously is more frequently and randomly in time and space position. That brings more challenges for radio monitoring system on multi-parameter signal recognition and flexible target tracking. With the advantage of flexibility and mobility, Unmanned Aerial Vehicle (UAV) assisted monitoring system is rapid advancing. In this paper, we propose a new framework for multi-parameter integrated recognition on OSS, including category and number of signals identification, position and transmitted power estimation, through UAV-assisted radio monitoring systems. Cooperative multi-layer Non-negative Matrix Factorization (NMF) recognition (CMNR) is firstly designed to solve the problem of OSS multi-parameters integrated recognition. To further reduce the computational complexity and time identically, joint multi-layer NMF recognition (JMNR) is then developed by fully using the data correlation among different monitoring nodes and significantly improve the integrated recognition accuracy based on joint matrix optimization. The flexibility and mobility of UAV monitoring nodes inspire the design of the centroid-seeking flying strategy (CSFS) for JMNR applications. In the simulation, the signal category recognition accuracy of JMNR and CMNR both achieve at least 27.1% higher than traditional blind signal separation (BSS) methods. In the large monitoring area, the position accuracy of JMNR even obtain 11.9% higher than that using CMNR. CSFS also exhibits 20.9% and 25.3% improvement on respective position and power estimation accuracy under large monitoring area and low height of UAVs conditions.