SINGLE POST

Elimination of Data Identification Problem for Data-Dependent Superimposed Training

In data-dependent superimposed training (DDST) schemes, the data-induced interference that occurs during channel estimation is eliminated at the expense of data distortion. Unfortunately, this distortion causes a data identification problem (DIP), where data sequence cannot be uniquely determined at the receiver end. Although DIP has been widely investigated in the literature, an effective solution for preventing its occurrence has yet to be proposed.

Accordingly, in contrast to previous studies, which explained the occurrence of DIP only for a special case, the present study explores the theoretical foundations for DIP in DDST schemes using a signal subspace technique and derives the general conditions under which DIP occurs. It is shown mathematically that the occurrence of DIP is related to the adopted modulation scheme and the pilot pattern. Therefore, a pilot design criterion that theoretically eliminates DIP is proposed. The simulation results show that the proposed pilot design successfully eliminates the error floor in the bit error rate performance of DDST schemes.