ELoran signal message recognition algorithm based on GTCN-transformer

Abstract

The Enhanced Long Range Navigation (eLoran) system serves as a crucial backup to the Global Navigation Satellite System (GNSS), leveraging advantages, such as low signal frequency, high transmitter power, and stable propagation distance. However, the prevailing demodulation techniques employed by the eLoran system, which are largely based on conventional digital signal processing, are susceptible to substantial inaccuracies when confronted with intense interference and complex environmental conditions. This paper introduces a novel GTCN-Transformer network designed for the specific task of recognising message in eLoran pulse group signal. The network is constructed by enhancing the architecture of Temporal Convolutional Networks (TCN) and integrating the Transformer mechanism. In order to extract significant features from the pulse group signal, a sequence dataset was obtained by using cepstral analysis. Subsequently, the GTCN-Transformer network is deployed to recognise the message contained within the eLoran pulse group signal. The experimental results demonstrate that the GTCN-Transformer network achieves a recognition accuracy of over 95% for eLoran signal message information when the SNR exceeds 10 dB, even in the presence of sky-wave and cross-interference signals. Moreover, a comparative analysis with recurrent neural network (RNN) reveals that the GTCN-Transformer network outperforms these architectures in terms of recognition accuracy.