SNR-Gated LSTM-Conditioned Diffusion Model for MIMO Channel Estimation
The method proposes an SNR-gated diffusion framework for MIMO channel estimation that uses time-series data and an LSTM to capture temporal patterns. It denoises channel estimates in the angular domain, using a special shortcut to balance observation accuracy at high signal levels with learned priors at low signal levels. This results in better accuracy than existing methods while maintaining low inference latency.