ResNeXt-LSTM Aided Sub-Nyquist Power Spectrum Sensing for Reconfigurable Intelligent Surface-Enhanced Orthogonal Frequency Division Multiplexing Wireless Communications

This paper presents ResNeXt-LSTM assisted sub-Nyquist spectrum sensing of orthogonal frequency division multiplexing (OFDM) signal in reconfigurable intelligent surface (RIS)-enhanced cognitive radio (CR) enabled wireless transmission scenario. ResNeXt is a state-of-the-art deep convolutional neural architecture consisting of multiple parallel transformation pathways, which combined with long short term memory (LSTM) can predict the spectrum occupancy status in CR-enabled wireless transmission with enhanced accuracy, compared to conventional convolutional recurrent neural architectures. Since sub-Nyquist sampling can alleviate the burden of extremely high sampling rate on the receiver, the performance of the spectrum sensing approach can be further improved by applying sub-Nyquist sampling to the received signal. Combination of sub-Nyquist sampling and ResNeXt-LSTM can predict the spectrum occupancy status with improved accuracy by reducing the reconstruction noise. Computer simulations demonstrate the effectiveness of the proposed approach in providing good performance under low signal-to-noise ratio (SNR) conditions, compared to conventional approaches.