Predicting CSI for Link Adaptation Employing Support Vector Regression for Channel Extrapolation

Link adaptation in LTE-A is based on channel state information (CSI). For time-selective channels, CSI might be out-dated already in the next subframe. Hence, CSI prediction must be employed. This paper investigates support vector regression (SVR) for channel extrapolation and prediction. SVR is applied for learning from the previous channel estimates in order to predict the CSI of the following ones. Simulation results show that the proposed method performs better than simple linear prediction methods and close to minimum mean square error prediction especially in a reasonable signal to noise ratio regime.