AI-Enhanced Channel Identification in 5G and Future Wireless Systems

This paper demonstrates the critical role of accurate channel distribution information in meeting the high-reliability, low-latency, and high-throughput requirements of 5G and forthcoming 6G networks for both military and civilian applications. A gap in the existing literature is identified: the joint learning of modulation and channel distribution information using a single model in a hybrid manner has not been explored. To address this gap, a novel convolutional neural network (CNN) architecture is proposed for the joint classification of BPSK-modulated signals transmitted over Rayleigh, Rician, and Nakagami fading channels. A custom MATLAB-based dataset comprising 50,000 I/Q samples is generated, and the model is trained in Python/Keras across SNR levels ranging from 0 to 20 dB. Performance evaluations reveal that classification accuracy increases from 41.6% at 0 dB to over 92% at 10 dB, reaching 97% at 20 dB; under high-SNR conditions, Nakagami and Rician channels are classified without error. These results demonstrate that the proposed hybrid CNN approach significantly improves channel identification performance under favorable noise conditions. By addressing modulation and channel classification simultaneously, this method is expected to pave the way for future extensions to MIMO systems, diverse modulation schemes, and validation with real-world measurement data.