Deep learning based channel estimation optimization in VLC systems

Abstract This paper aims to improve the channel estimation (CE) in indoor visible light communication system. The proposal of this deals with a system that depends on comparison between Deep Neural Network (DNN), Yolo v3, and Kalman filter (KF) algorithm, for two optical modulation techniques; asymmetrically clipped optical-orthogonal frequency-division multiplexing (ACO-OFDM) direct current frequency division (DCO-OFDM). CE can be evaluated by error rates received bits, where increased means performance decrease vice versa. Receiving less errors at receiver indicates improved performance. Hence, main aim our work is rate using different estimators. Furthermore, we apply automatic hyper-parameter approach Bayesian optimization, v3 model reduce positioning error. metric parameter bit (BER) determine improvement ratio systems. based training OFDM samples signal labels which are corresponding signals OFDM. At BER = 10 −3 DCO-OFDM, DNN outperforms KF 1.7 dB (8.09%) energy per noise $$(E_{b} {/}N_{o} )$$ ( E b / N o ) axis. Also, ACO-OFDM , achieves better results than about 1.9 (11.8%) ){ }$$ For values M QAM, average ~ 1.2 (~ 13%).