Super Capacity SRS Design for 5G and Beyond using Channel In-painting

Reliable communication of data in modern wireless systems requires accurate channel state information (CSI). Sounding Reference Signal (SRS) based CSI acquisition enables the estimation of the channel between the base station and user equipment through the uplink transmission of known SRS by the user equipment to the base station. However, limited SRS resources and limited Signal to Noise Ratio (SNR) coverage for which SRS-based CSI acquisition can be performed renders the acquisition of CSI challenging. We propose an approach to perform sparse non-uniform SRS resource allocation and use a masked auto-encoder with a vision transformer backbone to reconstruct full band channel from partial sub-band information. Experiments on data emulating transmission over CDL-A, B, and C channels demonstrate that the proposed method can achieve state-of-the-art normalized mean square error (NMSE) using only 25% of the band information. This translates to a four fold increase in SRS resource capacity and 6dB improvement in SRS coverage under current 5G NR specifications.