Channel order selection based on spectral matching for blind channel estimation

Second-order blind channel estimation of single-input multiple-output (SIMO) finite impulse response (FIR) channels is considered. Selection of channel order (model complexity) is a crucial part of blind channel estimation because true channel order is practically unknown, yet it considerably affects estimation results. In this contribution, usage of spectral matching criterion for channel order selection is discussed. The criterion is based solely on available channel output with mild assumption on the input. Simulation results show that the overall blind estimation performance can be significantly improved by the proposed order-selection criterion compared to existing ones. Furthermore, it also outperforms existing blind channel order methods claimed to be robust to overestimated channel order.