Data-Dependent Superimposed Training for Noncoherent Channels

This paper focuses on noncoherent coding design for fast fading channels, where the receiver does not have reliable channel state information. Unitary space-time codes and training based schemes have been proposed historically. We propose a data-dependent superimposed training scheme to improve the performance of training based codes. The transmitter is equipped with multiple training sequences and dynamically selects a training sequence for each data sequence to minimize channel estimation error. The set of training sequences are optimized to minimize pairwise error probability between codewords.