Classification of RF transmitters in the presence of multipath effects using CNN-LSTM

Radio frequency (RF) communication systems are the backbone of many intelligent transport and aerospace operations, ensuring safety, connectivity, and efficiency. Accurate classification of RF transmitters is vital to achieve safe and reliable functioning in various operational contexts. One challenge in RF classification lies in data drifting, which is particularly prevalent due to atmospheric and multipath effects. This paper provides a convolutional neural network based long short-term memory (CNN-LSTM) framework to classify the RF emitters in drift environments. We first simulate popular-used RF transmitters and capture the RF signatures, while considering both power amplifier dynamic imperfections and the multipath effects through wireless channel models for data drifting. To mitigate data drift, we extract the scattering coefficient and approximate entropy, and incorpo-rate them with the in-phase quadrature (I/Q) signals as the input to the CNN-LSTM classifier. This adaptive approach enables the model to adjust to environmental variations, ensuring sustained accuracy. Simulation results show the accuracy performance of the proposed CNN-LSTM classifier, which achieves an overall 91.11% in the presence of different multipath effects, bolstering the resilience and precision of realistic classification systems over state of the art ensemble voting approaches.