A Light-Fidelity Machine Learning Assisted Hybrid MIMO-OFDM Framework for Efficient Channel Estimation and Detection in Indoor Li-Fi Systems

Abstract

This paper presents a novel light-fidelity machine learning (ML)-assisted framework for improving channel estimation and signal detection in hybrid MIMO-OFDM-based indoor Li-Fi communication systems. Conventional estimation techniques such as Least Squares (LS) and Minimum Mean Square Error (MMSE) suffer from a trade-off between computational complexity and estimation accuracy, especially under dynamic channel conditions and low Signalto-Noise Ratio (SNR). To address these challenges, this work introduces a hybrid approach that integrates classical signal processing with a lightweight ML model for error refinement. Unlike deep learning approaches, the proposed method utilizes a regression-based model to enhance LS estimation with minimal computational overhead. The system is evaluated using Bit Error Rate (BER), Mean Square Error (MSE), and throughput metrics under varying SNR conditions. Simulation results demonstrate that the proposed framework achieves near-MMSE performance while maintaining low complexity, making it suitable for real-time indoor Li-Fi applications. The proposed model also reduces pilot overhead and improves spectral efficiency, highlighting its potential for next-generation wireless communication systems.