CHANNEL STATE INFORMATION RECONSTRUCTION FROM SPARSE DATA

摘要

Accurate downlink channe! estimates are catenated based on infrequently transmitted Channel State information (CS!) feedback data from a UE 20. A plurality of non-uniformly spaced digital CSI feedback samples, representing the frequency response of a downlink communication channel, is received from the UE, The received CS! feedback samples are demodulated and inverse quantized. A time domain tap deiay channel model is generated from the inverse quantized CSI feedback samples. The time domain tap delay channe! model may be frequency-transformed to obtain a reconstructed frequency response of the downlink communication channel in the frequency domain. Alternatively, channel delays may be estimated based on prior delays and/or known references signals transmitted on the uplink. Channel estimates between CS! reporting instances may be predicted, such as by a sample & hold or a linear predictor. The delays may be presumed fixed, and Kalman filter coefficients evolved over time.