Multi-scale Feature Extraction Channel Equalization Method Based on Res2net Module

Under the condition of time-varying multi-path channel, to address traditional channel equalization methods cannot effectively eliminate Inter Symbol Interference, a multi-scale feature extraction channel equalization method based on Res2net module is proposed. The method adopts Res2net network technology, firstly, multi-scale feature extraction is carried out in the high dimensional space for the components of the received signal that are subject to ISI, then the reconstruction of the interference signal is completed by the extracted multi-scale features, and finally, based on the idea of residual difference, the reconstructed interference signal is used to correct and compensate the received signal in order to achieve the channel equalization, which reduces the bit error rate of the communication system and improves the reliability of the transmission of the communication system. Simulation results show that, compared with the traditional channel equalization method, the proposed method reduces the bit error rate by 6% in the Gaussian white noise channel environment and 3% in the colored noise channel environment under the multi-path Rayleigh channel.