Robust Deep Blind Equalization Network Under Time-Varying Channels

As a key technology to recover transmitted symbols without any prior information of the incoming signals, blind equalization (BE) has gained widespread attention from researchers. However, conventional BE algorithms often assume a constant or slowly varying channel, limiting their effectiveness in rapidly changing environments. In this letter, we propose a transformer-based deep BE network specifically designed for time-varying channels. To compensate for carrier frequency offsets (CFOs), we incorporate an adaptive parameter correction module. Moreover, the network implements a pre-extractor to capture multi-scale channel features, enabling it to handle time-varying conditions effectively. Simulations demonstrate that the proposed method achieves performance comparable to orthogonal frequency division multiplexing (OFDM) systems while significantly reducing pilot overhead.