Deep Learning-Assisted OOK Detection in OFDM-MIMO Systems for Robust Wireless Communication

Abstract

Orthogonal Frequency Division Multiplexing (OFDM) combined with Multiple Input Multiple Output (MIMO) has become a fundamental architecture in modern wireless communication systems due to its high spectral efficiency and robustness against multipath fading. However, reliable signal detection in such systems remains challenging, particularly when simple modulation schemes such as On-Off Keying (OOK) are employed. OOK offers low complexity and energy efficiency but suffers from high sensitivity to noise, interference, and channel impairments. This paper proposes a machine learning-assisted detection framework for OOK in OFDM-MIMO systems to enhance detection accuracy while maintaining computational efficiency. The proposed approach integrates feature extraction, signal preprocessing, and supervised learning-based classification to improve symbol detection under realistic channel conditions, including Rayleigh fading and additive noise. Simulation results demonstrate that the proposed model achieves improved Bit Error Rate (BER) performance, reducing BER from 0.08 to 0.04, while achieving an accuracy of 94.8%. Additionally, the system exhibits improved robustness against inter-symbol interference and multi-antenna interference. The results confirm that machine learning-based detection provides a practical and scalable solution for next-generation wireless communication systems.