Nonintersecting Subspace OSTBCs for Blind ML Detection

Recently there has been growing interest in employing the orthogonal space-time block codes (OSTBCs) for blind maximum-likelihood (ML) detection. Several independent works have suggested that OSTBCs are favorable space-time codes from a blind receiver implementation standpoint. In this work we turn our attention to blind ML identifiability, with an emphasis on a special class of codes called the nonintersecting subspace (NIS) OSTBCs. We show a powerful property that NIS-OSTBCs are uniquely identifiable up to a sign for any nonzero channel. However, many existing OSTBCs are not NIS. We propose a code construction procedure that can convert an existing OSTBC to an NIS-OSTBC. Simulation result are provided to support our theoretical findings.