Fast Blind Channel Equalization Based on Online Deep Neural Network

Blind equalization has recently attracted much attention due to its exceptional potential for enhancing spectrum efficiency in wireless communication systems. However, the error performance of existing methods often fails to meet expectations with short run time and sequence lengths. Hence, this letter proposes a low-complexity online deep neural network-based algorithm to optimize the blind equalization problem to address these limitations. Specifically, by transforming the classic symbol-based decision (SBD) method into a low-complexity neural network, we quickly estimate the optimal equalizer coefficients for each signal sequence through online training. Indeed, the simulation results demonstrate that the proposed method outperforms conventional state-of-the-art methods in terms of running speed and bit error rate (BER) performance, while only a short-length signal sequence is required.