Channel-Attention-Enhanced LSTM Neural Network Decoder and Equalizer for RSE-Based Optical Camera Communications

In an RGB-LED-based optical camera communication system, it is essential goal to have better performance in the data rate and BER. However, a higher symbol rate, due conventional sampling algorithm, deterioration of transmission brought by inter-symbol interference inter-channel significant. Innovatively, this paper, sub-image obtained captured frame received video encoded channel-attention-Net-based encoder generate descriptor without existing methods. Moreover, we propose LSTM-based equalizer decode mitigate deterioration. Utilizing long-short-term memory LSTM unit, not only can reduce bit error rates but also increase rate. The experimental results show that at 46 kbaud/s, record-high 44.03 kbit/s achieved under random while still meeting pre-forward correction requirement.