Dual-labelled multi-Bernoulli filter based on specific emitter identification

In complex electromagnetic environments, airborne passive bistatic radar encounters the challenge of associating emitters with measurements for multi-target tracking. The authors propose a solution based on specific emitter identification technology. Firstly, generative adversarial networks (GANs) are utilised to extract and classify emitter signals using radio frequency fingerprint (RFF) features. The classification results are then used to construct a set of emitter labels for pre-processing the measurement data. Subsequently, the pre-processed measurement data set is input into the labelled multi-Bernoulli filter framework, which is extended to a dual-labelled (target label and emitter label) multi-Bernoulli filter. This filter jointly predicts and updates the multi-target posterior density, enabling the estimation of multi-target trajectories. The effectiveness of the proposed algorithm is validated using two experiments. The results demonstrate that the GAN based on RFF features effectively identifies emitter signals. Moreover, the dual-labelled multi-Bernoulli filter, based on specific emitter identification, accurately estimates multi-target trajectories using measurement data from an airborne passive radar of the multi-transmit single-receive type. This approach provides a novel and effective solution to the multi-target tracking problem in complex electromagnetic environments.