Quantized Channel Prediction And Subspace Tracking Algorithms For Time-Varying TDD-MIMO Systems

By using multiple antennas at both transmitter and receiver, MIMO (Multiple-Input Multiple-Output) technique has the ability to improve system capacity significantly without increasing bandwidth or transmit power. In the TDD-MIMO systems, uplink channel station information (CSI) and downlink CSI are reciprocal, so base station can use uplink CSI for downlink transmission. However, time-varying channel will lead to channel non-reciprocity, which degrades system capacity. This thesis addresses channel prediction for compensating non-reciprocity in TDD MIMO systems. As follows are the main contributions:1. This thesis analyzes the effect of time-varying channel on the channel reciprocity in TDD MIMO systems, and then verify it through simulations and the derivation capacity formulas.2. A method based on quantization is proposed. The eigenvector matrix obtained by singular value decomposition (SVD) on channel matrix is quantified by the Lloyd algorithm and then is predicted by the LS-SVM algorithm. Simulation results show that the proposed method can effectively compensate the channel non-reciprocity and improve system capacity.3. The subspace tracking algorithm based on the LS-SVM is proposed. The changes of the predicted channel transfer matrix in time-varying channels are tracked, and then used for calculating channel response. Simulation results show that the algorithm can improve the accuracy of channel prediction and decrease bit error rate (BER).