A new spectrum sharing protocol with simultaneous wireless information and power transfer (SWIPT) is proposed to cope with the increasingly prominent problem of spectrum and energy scarcity. It operates within a cognitive radio network (CRN) in the context of cellular IoV (C‐IoV), enabling bidirectional communication between two vehicles parked within the base station coverage (VnBSs) while facilitating cooperation for a pair of primary users (PUs), i.e., VnBSs can act as relays to provide cooperation communication for the cell–edge vehicle user (eVU). Unlike most existing work, both VnBSs can use time switching (TS) to obtain energy from radio frequency (RF) signals emitted from the base station. In order to enhance the reliability of the network, this study incorporates the automatic repeat request (ARQ) technique in the CRN supported by the nonorthogonal multiple access (NOMA) and SWIPT, which has not been performed in other works. Based on this, the transmission is divided into one energy harvesting (EH) phase and three information processing (IP) phases. A new packet for PUs is transmitted in the first IP phase and is allowed to be retransmitted twice in the last two IP phases depending on the decoding. VnBSs act as relays to obtain energy in the EH phase to assist in retransmitting the PU’s packets and sending their own packets in the last two IP phases. The system states are analyzed by building a one‐dimensional Markov chain, and the end‐to‐end outage probability (OP) is calculated for each state under the Nakagami‐m fading channel. Using these two results, the OP of the primary and secondary networks, system throughput and energy efficiency (EE) are derived. Finally, the validity of the derived results is verified by Monte Carlo simulation using MATLAB and compared with the protocol without ARQ, and the protocol proposed shows a better performance.

In many practical communication environments, traditional feature extraction methods in spectrum sensing fail to fully exploit the information of primary users. Additionally, conventional machine learning methods have weak learning capabilities, making it difficult to maintain efficient and stable spectrum sensing performance in complex noise environments. Furthermore, non‐Gaussian noise can significantly affect the detection performance of spectrum sensing. To address these issues, this paper first proposes a feature extraction method based on Hierarchical Fuzzy Dispersion Entropy (HFDE) to better extract high‐frequency and low‐frequency information from signal samples, providing more comprehensive features for subsequent models to optimize feature extraction effectiveness. Then, a parallel model combining Convolutional Neural Networks (CNN) with Gated Recurrent Units (GRU) is constructed to enhance learning ability. While CNN extracts local features, GRU processes temporal relationships, and the features output by both are concatenated to achieve effective feature learning and temporal modeling of primary user signal data represented by HFDE. Finally, using the feature vectors output by the CNN‐GRU model, detection statistics and detection thresholds for spectrum sensing are constructed for online detection. Simulation results validate the effectiveness and robustness of this method in spectrum sensing under non‐Gaussian noise. In the presence of significant non‐Gaussian noise intensity and a signal‐to‐noise ratio of −14 dB, the detection probability can reach 97.1%. Additionally, for the detection of unknown signals, the model can still maintain a detection probability of over 90%.