Communication Emitter Identification Under Asynchronous Acquisition Condition

An asynchronous data acquisition method for communication emitter identification under non-cooperative conditions is proposed to alleviate the problem of recognition drift. Firstly， a measurement signal model is established under asynchronous acquisitions. The impact of time-frequency asynchronous scenarios， including multipath channel interference， relative motion of the target， and receiver-specific differences， on the measurement signal is analyzed in detail. Time delay and frequency shift are demonstrated to be the primary factors contributing to the divergence of identification features. Secondly， to mitigate the interference of time delays and frequency shifts on the subtle characteristics of the signal， these effects are transformed into coordinate positions of signal features on time-frequency domain （TFD） using short-time Fourier transform （STFT）. A unified time-frequency resolution scale is applied to enhance the stability of signal features. Measurement errors in TFD are derived and analyzed， which highlight that time delay and frequency shift induce position jitter in time-frequency features （TFFs）， and thus， indirectly affect feature measurement accuracy. Thirdly， to reduce the influence of TFFs position jitter， refined time-frequency measurement technology is employed to extract fine-grained signal features. An algorithm to eliminate the time-frequency fence effect based on frequency-domain frequency shift compensation is presented. Finally， leveraging the refined TFFs of the signal， emitter individual identification is reformulated as a fine-grained image recognition problem based on deep learning. The effectiveness of the proposed algorithm is validated using the DenseNet201 transfer learning network and image enhancement technique. Field experiments demonstrate that for shipborne automatic identification systems （AIS） of the same type， under asynchronous acquisition conditions with varying sampling devices， parameters， time and space， the Top-1 individual identification accuracy rate for 10 communication emitters exceeds 85%.