MMSE-optimal training sequences for spectrally-efficient Multi-User MIMO-OFDM systems

This paper proposes a new family of optimal training sequences in terms of minimizing the mean-square channel estimation error for spectrally-efficient Multi-User MIMO-OFDM systems with an arbitrary number of transmit antennas and an arbitrary number of training symbols. It addresses uplink transmission scenarios where the users overlap in time and frequency and are separated using spatial processing at the base station. In particular, optimal training sequences can be obtained easily from standard signal constellations such as QPSK with desired low PAPR, making it appealing for practical use.