MMSE Training Design for Filter Bank Multicarrier Systems with Per-Subcarrier Channel Estimation

Filter bank based multicarrier (FBMC) systems present an alternative solution to cyclic prefix based orthogonal frequency division multiplexing (CP-OFDM) in wireless environments with multipath propagation. In this contribution we propose a novel training sequences design method to be employed in a per-subcarrier maximum likelihood (ML) narrowband channel estimation scheme, recently developed by the authors. In the training design we take into account the knowledge of transmitter and receiver side prototype filters to calculate the pilot sequences. The method targets at the minimum mean squared error (MMSE) of the channel estimation and uses a gradient projection algorithm to find the optimum sequences. In the numerical results we show a gain of up to 4 dB for the same MMSE value in a low to medium E
_b
/N
₀
regime.