Channel estimation for MIMO-OFDM with training overhead trade-off

We propose a new technique for initial channel estimation in packet-mode MIMO-OFDM. A maximum likelihood estimator uses observations of the channel spectra obtained by means of a carefully designed multi-transmitter channel probing sequence. The sequence length is controlled by two competing parameters, which allow for an explicit and fine trade-off between the channel estimation overhead and the quality of the resulting estimates. Compared to the training sequence used in the IEEE 802.11a standard, our technique requires 6 to 8 times less training symbols in order to attain channel estimates of equal accuracy.