Bayesian Learning-Based Sparse Channel Estimation in Visible Light ADO-OFDM Systems

This paper presents an innovative scheme for estimating the channel impulse response (CIR) in sparse multipath conditions for asymmetrically clipped direct current-biased optical OFDM (ADO-OFDM) visible light communication (VLC) systems, utilizing Bayesian learning (BL) techniques. We derive a multipath CIR model capturing both specular and diffusive reflections within the VLC system. Subsequently, we present a novel scheme for estimating the CIR in sparse multipath scenarios using the BL paradigm, which leverages the inherent sparsity of the multipath CIR in the delay domain. This scheme necessitates a constrained set of pilot subcarriers, thereby reducing pilot overhead when juxtaposed with traditional state-of-the-art channel estimation (CE) techniques. To assess the performance of the proposed BL-based paradigm for estimation, we compute the Oracle-MMSE (O-MMSE) along with the Bayesian Cramer Rao lower bound (BCRLB). Our extensive simulations reveal that even with a lower pilot overhead, the suggested BL method surpasses other conventional and sparse CE techniques across key metrics such as bit error-rate (BER) and normalized mean-square-error (NMSE).