Channel Equalizer Research Based on Deep Neural Networks

In digital communication, data transmission is impacted by channel distortion. Inter-Symbol Interference (ISI) arises from the dispersive characteristics of the channel. To enhance reliability and support high-speed communication, channel equalization techniques are employed at the receiver to mitigate the effects of ISI. This paper proposes an effective equalizer, named CLE, which combines Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks. As the terrestrial wireless channel is mainly affected by small-scale fading such as multipath fading and Doppler shift, the Rayleigh fading channel is modelled using the Jakes approach. Compared with traditional Decision-Feedback Equalizer (DFE), CLE can achieve the performance improvement of close to 4~6dB on the premise of reaching the same bit error rate. The experiment results show that CLE is more efficient and less complex than the equalizer with similar techniques, resulting in faster processing and reduced computational demands, making it more practical for real-world applications. It means that our method is better equipped to handle the fluctuations in channel conditions, maintaining reliable performance even as the environment changes.