Semi-Blind Channel Estimation for Frequency-Selective MIMO Systems

The task of channel estimation in a frequencyselective multi-input multi-output (MIMO) system presents a real challenge for conventional training-based solutions, since the huge number of channel coefficients demand a huge number of training symbols, which will significantly reduce the system bandwidth-efficiency. As a possible countermeasure, we propose a semi-blind channel estimation algorithm, which can learn channel coefficients accurately when only a short training is available. The main idea is to use blind data/channel estimation algorithms to improve the quality of the channel knowledge obtained from the short training. With the proposed semi-blind algorithm, the required training length will no longer be unaffordable for frequency-selective MIMO systems.