CNN–LSTM-assisted joint channel estimation and signal detection for optical wireless communications

The performance of signal detection in a multi-input multi-output (MIMO) orthogonal frequency division multiplexing (OFDM) optical wireless communication (OWC) system is significantly affected by channel fading and atmospheric turbulence. Accurate channel state information (CSI) is crucial for reliable channel estimation, but obtaining it typically requires a high pilot overhead. To address this issue, we propose a joint channel estimation and signal detection (JCESD) algorithm that reduces the pilot overhead while improving the detection performance with a moderate increase in computational complexity. By leveraging convolutional neural network (CNN) and long short-term memory (LSTM) networks, the JCESD algorithm learns the mapping relationship between atmospheric turbulence channels and signals, thereby mitigating the challenges associated with high-complexity signal processing at the receiver. Additionally, layer normalization is applied during data pre-processing to enhance overall performance. Simulation results demonstrate that the proposed JCESD algorithm outperforms the LS-ZF algorithm and achieves an approximate 1 dB gain over the LS-MMSE algorithm under weak turbulence conditions at a bit error rate (BER) of 10^-5 in the 2×2 system when using eight pilots, with a 0.9 dB gain when using four pilots-highlighting its capability to maintain superior performance while reducing pilot overhead. Additionally, the JCESD algorithm exhibits superior robustness in atmospheric turbulence channels while effectively reducing pilot overhead and maintaining manageable computational complexity.