Dynamic Autoencoder-Based Framework for Performance Enhancement in OFDM Systems Using CNN and Attention Mechanisms

Traditional orthogonal frequency division multiplexing (OFDM) systems face significant performance degradation under dy￾namic channel conditions due to fixed channel estimation and static transmission parameters. Existing autoencoder-based
OFDM models improve end-to-end learning but lack adaptive mechanisms to handle varying SNR and multipath fading effec￾tively. In this manuscript, dynamic autoencoder-based framework for performance enhancement in OFDM systems using CNN
and attention mechanisms (DAF-OFDM-CNN) is proposed. This paper proposes a dynamic autoencoder-based framework to
enhance the performance and reliability of OFDM systems under fluctuating signal-to-noise ratio (SNR) conditions. The frame￾work integrates convolutional neural networks (CNN) and an attention mechanism within autoencoder architecture to improve
feature extraction, channel estimation, and adaptive transmission. The system dynamically prioritizes important signal features,
enabling effective data recovery and robust communication in multipath fading environments. Simulation results demonstrate
that the proposed model significantly improves error rates, throughput, and latency compared to conventional OFDM systems,
confirming its potential for next-generation wireless communication networks.