Study Of ICA Based Blind Channel Identification In MIMO OFDM Systems

An accurate channel state information (CSI) is crucial for the performance of multiple input multiple output orthogonal frequency division multiplexing (MIMO OFDM) systems. Usually it can be obtained by sending training data and pilot tones. The training overhead required, unfortunately, is more significant in estimating multiple input multiple output channels. To avoid this problem, the blind channel identification attracts more interest.In this paper, two ICA based blind channel identification algorithms in MIMO OFDM systems are proposed. One is the ICA-only algorithm, the other is the ICA-MMSE algorithm. At present, the ICA based blind channel identification algorithm employs ICA only in one subcarrier and unmixes the remaining subcarriers using a minimum mean square error (MMSE) approach. However, this method results in severe error propagations over subcarriers as the subcarrier being unmixed uses the stream estimates from the previous subcarrier as a conference. In this paper, we use ICA in all subcarriers to obtain the channel parameters. The ICA-only method proposed in this paper obtains better performance by using ICA in every subcarrier. But since a large number of subcarriers may be used in an OFDM system and additionally a reordering step is required in each subcarrier to overcome the scaling and order indeterminacies of ICA, this receiver has a high complexity. Then the second algorithm is proposed, which uses ICA in one subcarrier and uses the same reference subcarrier for unmixing all the remaining subcarriers with the MMSE method. Theory analysis and experimental results show that this algorithm has good performance in channel identification as well as low complexity.