فهرست نشریات علمی دسترسی آزاد ایران

The rapid expansion of 5G and upcoming 6G networks requires wireless image transmission schemes that deliver ultra-reliable, low-latency, and energy-efficient performance over highly dynamic multiuser channels. This paper presents a deep convolutional neural networks (DCNN) DenseNet- with Enhanced Marine Predator Algorithm (EMPA) and a hybrid deep learning and metaheuristic framework for wireless image transmission in non-orthogonal multiple access (NOMA)-aided massive multiple-input multiple-output (MIMO) systems. The proposed model combines a dense convolutional network (DenseNet) encoder-decoder with an enhanced marine predator algorithm (EMPA) optimizer to jointly tune channel coding power allocation, and resource scheduling under multi-user interference. DenseNet model dense feature propagation efficiently extracts multi-scale image representations, whereas EMPA adaptively minimizes transmission error and latency by optimizing NOMA power coefficients and MIMO beamforming weights. The benchmark proposed model DenseNet-EMPA against recent hybrid learning optimization paradigms, including a comparison of exit models such as deep reinforcement learning with self-optimized particle swarm joint source-channel coding (DRL-SOPSOJSCC), enhanced golden optimization algorithm integrated with deep reinforcement learning (EGOA-DRL), and improved golden jackal optimization with Bi-LSTM capsule network (IGJO-BiLSTM-CapsNet). Simulation results under realistic 5G/6G-orientated parameters show that DenseNet-EMPA achieves up to 18–24% lower bit error rate, 15% higher spectral efficiency, and 20% reduced latency compared to state-of-the-art approaches, while maintaining robust performance across varying SNR, user density, and mobility profiles. The performance results demonstrate that combining deep convolutional encoders with bio-inspired metaheuristics provides a powerful design paradigm for next-generation wireless adaptive image transmissions.