Optimised time-frequency analysis for drone detection in a multi-signal environment

Abstract

The abuse of recreational drones has caused security issues, crimes, and privacy problems. Thus, drone detection is needed as evidence for law enforcement or support countermeasures such as radio jamming, and remote control override. One way to detect drones is by monitoring the radio frequency (RF) link between the remote control and drone. The challenge is that recreational drones operate at the 2.4 GHz and 5.8 GHz industrial, scientific and medical (ISM) bands, which have wide bandwidth of 100 MHz and 150 MHz respectively make the analysis costly. Furthermore, the drone signals like frequency-hopping spread spectrum (FHSS) and hybrid spread spectrum (HSS) are time-varying and require a time-frequency analysis (TFA). Also, the choice of window size is crucial for a TFA due to the uncertainty principle in time-frequency representation (TFR). Moreover, other wireless technologies in the environment, such as direct sequence spread spectrum (DSSS) and orthogonal frequency-division multiplexing (OFDM), which is Wi-Fi operates in the same ISM band could interfere with the drone signal detection activity. In this thesis, an adaptive stepped frequency scan spectrogram (Adaptive-SFSS) was developed to analyse a large bandwidth at a lower sampling rate, including an adaptive window size estimation. In the Adaptive-SFSS, the received signal is divided into multiple sub-bands and scan through the large analysis bandwidth, the window size is estimated by balancing time and frequency resolution, the channel frequency and hop duration are estimated from TFR and used to derive the instantaneous frequency (IF). Three types of drone signals, the fast FHSS, slow FHSS, and HSS, together with two types of background signal, the DSSS and Wi-Fi were simulated. Then, the simulated received signal was analysed by the Adaptive-SFSS and compared with the adaptive wideband spectrogram (Adaptive-WS), the non-adaptive SFSS and WS. The performance of the Adaptive-SFSS was verified by Monte-Carlo simulation with 20 realizations at a signal-to-noise ratio (SNR) range from -1 6 dB to 12 dB. In the presence of additive white Gaussian noise (AWGN), the Adaptive-SFSS obtained a detection cut-off point of -1 2 dB for fast and slow FHSS and -5 dB for HSS. Additional background signals such as DSSS and Wi-Fi increased the cut-off point to 5 dB for fast-FHSS, 7 dB for slow-FHSS, and 8 dB for HSS. The Adaptive-SFSS is better because it has a similar cut-off point as the WS even the sampling rate is 4 times lower and capable of choosing the right window size automatically, rather than trial-and-error which is the conventional way.