| Abstrakt: |
The non-linearity of Light Emitting Diodes (LEDs) has limited the efficiency of visible Light Communication (VLC) in terms of Bit Error Rate (BER). In this report, we propose a model-driven Deep Learning (DL) strategy for Optical Orthogonal Frequency Division Multiplexing (O-OFDM)-based VLC processes, that is employed as the Auto Encoder (AE) network system to minimize the LED non-linearity. The suggested system successfully includes communication domain knowledge into the design of the training cost function and network architecture, as opposed to the conventional fully computer-controlled autoencoder. Deep Recurrent Neural Network (Deep RNN) and Inverse Fast Fourier Transform (IFFT) are used at the emitter end to convert the binary data first into challenging I-Q codes for each O- OFDM sub-band. Then, for nonlinearity compensation and signal detection at the receiver, the symbol de-mapping is done and the demodulation is performed through a Deep RNN. DeepRNN's hidden layers are fine-tuned using the MMRFA algorithm for enhanced performance and efficiency. The proposed methodology performance is compared with the existing methods using performance metrics like BER, MSE, ACF, Power consumption, and Energy efficiency. The proposed solution outperforms several current approaches in terms of BER performances and improves operating duration, demonstrating the practicality and promise of DL in the VLC platform. [ABSTRACT FROM AUTHOR] |
Full text from SpringerLink