Hybrid Autoencoder-Based Machine Learning Framework for PAPR Reduction in OFDM Systems

Abstract

Orthogonal Frequency Division Multiplexing (OFDM) is widely employed in modern wireless communication systems due to its high spectral efficiency and robustness against multipath fading. However, a major limitation of OFDM systems is the high Peak-to-Average Power Ratio (PAPR), which leads to nonlinear distortion in power amplifiers and degrades overall system performance. This paper proposes a hybrid machine learning-based framework for efficient PAPR reduction using an autoencoder-assisted optimization approach. The proposed method integrates deep learning with classical signal processing to minimize signal peaks while preserving signal integrity. Experimental results demonstrate that the proposed model achieves a significant reduction in PAPR, from approximately 11.8 dB to 6.2 dB, compared to conventional techniques. Additionally, the Bit Error Rate (BER) is improved from 0.12 to 0.04, while maintaining a high prediction accuracy of 94.8%. The model also shows faster convergence with reduced loss values, reaching below 0.015 in final epochs. These results confirm that the proposed framework provides an effective balance between performance and computational efficiency, making it suitable for real-time wireless communication systems such as 5G and beyond.