A low complexity LMMSE channel estimation for OFDM systems

Abstract

In general, coherent orthogonal frequency division multiplexing (OFDM) detection requires knowledge about the reference amplitudes and phases of the constellation on each subcarrier. To achieve this, several channel estimation schemes have been investigated. In this thesis, we briefly introduce some techniques to estimate the channel in OFDM system. The least square (LS) algorithm is simple and has low complexity but the performance is not suitable for practical case. The linear minimum mean square error (LMMSE) channel estimator which makes full use of the correlation of the channel frequency response has good performance but it requires knowledge of the channel frequency correlation and the operating SNR. Also the complexity is so high due to the fact that it needs matrix inversion of autocorrelation matrix and the large number of complex multiplications. Although the computational complexity of the LMMSE estimator can be reduced by using singular value decomposition (SVD), LMMSE estimate is not acceptable because the complexity by the SVD cannot be negligible. Therefore, we propose a low complexity LMMSE channel estimator for OFDM system. We partition channel correlation matrix in frequency domain into small sub-matrices by using non-overlap and overlap techniques. The size of sub-matrix is determined by channel coherence bandwidth which defines a range of highly correlated channel frequency response. In case of non-overlap, we observe that mean square error (MSE) is higher at edge subcarriers of partitioned sub-matrix than other subcarriers. To compensate such a high MSE at edge subcarrier, we propose the overlap method. By utilizing overlap technique, we show that the MSE of the estimator can be reduced as compared with that using non-overlap method. Also the characteristic that the channel impulse response is independent for different paths in time domain is employed for diminishing both computational complexity and memory size of proposed estimators. F...