A Computationally Efficient Block Transmission Scheme Based on Approximated Cholesky Factors

Least square and minimum mean square error data estimation based on rectangular and circulant channel submatrices is compared in terms of estimation quality, storage requirements and the number of multiplications. The circulant case is applied in OFDM based transmission systems. When applying an approximated Cholesky decomposition, the computation of the estimator in the rectangular case may require even less computations than in the circulant case, while yielding better bit error rates in a time dispersive wireless transmission scenario. The considered approximations may improve the condition of the estimator and thus may even yield lower bit error rates than the exact estimator. The actual computation of the estimates is usually more efficient in the circulant case than in the rectangular.