| In recent years,with the increasing demand for wireless communication,people have put forward higher requirements for high-speed data transmission and security.The existing wireless spectrum resources cannot further meet the future demand,and Visible Light Communication（VLC）has emerged as a spectrum-rich,green and secure communication technology in the public eye,and has become one of the core technologies in the 6G communication architecture.When the signal is transmitted in the visible light communication system,it will be damaged by the environment and communication devices.Common damages include linear damage caused by intersymbol interference and nonlinear damage caused by device characteristics.In the process of high-speed signal propagation,nonlinear damage will seriously affect the transmission quality of the signal,resulting in an increase in the bit error rate between systems.In view of the problem of signal damage in visible light communication system,this thesis constructs a digital signal processing algorithm based on deep learning to compensate distorted signals,and its specific content is as follows:1.An Entity Extraction Neural Networks（EXNN）is proposed for visible light communication systems to compensate for damaged signals.First,the long-range extraction of the noisy signal is performed using a dilated convolution layer for the linear and nonlinear impairment signals present in the received signal.Secondly,the network of the residual structure is constructed to perform the subtraction operation on the separated noise and the received signal to complete the noise removal from the received signal.Finally,the signal after noise removal is processed using regression to ensure that the maximum compensation of the damaged signal is completed.In this thesis,the performance of EXNN network is verified by experimenting on Pulse Amplitude Modulation（PAM）system.Meanwhile,the EXNN algorithm is compared with the linear blind equalization algorithm and the Volterra-based nonlinear equalization algorithm under the restriction of the Hard Decision Forward Error Correction（HD-FEC）BER threshold value of 3.8×10-3.The experimental results show that the EXNN network improves the PAM The experimental results show that the EXNN network improves the Q-factor of the PAM and VLC system by 0.36 d B and1.57 d B,respectively.2.A Channel Attention Neural Network（CANN）equalizer is proposed to equalize the impaired signals in the Carrier-less Amplitude and Phase Modulation（CAP）system.Network equalizer is a special equalizer designed for CAP VLC systems,which compensates the signals of the in-phase and quadrature branches of the CAP modulation system,respectively.The CANN network equalizer is a special equalizer designed for the CAP VLC system,which compensates the signals of the in-phase and quadrature branches of the CAP modulation system.The second part uses the attention mechanism to filter the extracted features,retaining the important ones and discarding some useless ones,to finally achieve the compensation of distorted signals.In the experimental environment,the constructed CANN network equalizer is compared with the classical digital signal processing algorithm,and the experimental results show that the algorithm can effectively suppress the distortion of the system and reduce the BER of the signal at the receiver side by significantly compensating for the impairment in the signal.Compared with the linear blind equalization algorithm,the CANN equalization algorithm reduces the bit error rate by about 45.2%,and reduces the bit error rate by about 11.4%compared with the nonlinear equalization algorithm.In this thesis,the signal damage in the visible light communication system is equalized by the deep learning algorithm,and the effectiveness of the proposed deep learning algorithm is verified in the experiment,which provides a solution for the signal compensation in the visible light communication system. |
