Learning-Assisted Receiver for ACO-OFDM with Device Imperfections

• Published in: 2024 IEEE/CIC International Conference on Communications in China (ICCC) pp. 2012 - 2016
• Main Authors: Li, Li, Liu, Han, Mao, Tianqi, He, Dongxuan, Hou, Huazhou
• Format: Conference Proceeding
• Language: English
• Published: IEEE 07-08-2024
• Subjects: ACO-OFDM; convolutional neural network; device imperfections; Distortion; Fast Fourier transforms; Light emitting diodes; Performance evaluation; Quantization (signal); Receivers; Simulation; Visible light communication
• DOI: 10.1109/ICCC62479.2024.10681817

Visible light communication (VLC) has been regarded as an emerging technology to satisfy the ever-increasing demand of ultra-high-speed wireless communications. To guarantee the transmission efficiency, asymmetrically clipped optical-orthogonal frequency division multiplexing (ACO-OFDM) has been adopted. However, adversely effected by the device imperfections, which include the nonlinearity of light emitting diode and low-resolution quantization of analog-to-digital converter, the demodulation performance of ACO-OFDM receiver is limited. To tackle this problem, a learning-assisted receiver is designed, where convolutional neural network (CNN) is adopted to demodulate the received signal with distortion. More specifically, the received signal before fast Fourier transform (FFT) is input into the convolutional layer, which is helpful to exploit the signal feature even under device imperfections. Then, the demodulation is modeled as a classification problem, where the output of CNN is the demodulation likelihood information. Simulation results show that our proposed CNN can recovery information from the distorted signal, and improves the demodulation performance significantly.