Optimal Low-Complexity Orthogonal Block Based Detection of OTFS for Low-Dispersion Channels

Orthogonal time frequency space (OTFS) modulation constitutes a promising technology for high-mobility scenarios. However, the detection of OTFS systems imposes substantial complexity. Hence, we propose a novel orthogonal block (OB) based detection scheme for significantly reducing the OTFS detection complexity without any performance loss with integer Doppler shifts. This is achieved by recognizing that the received signal can be partitioned into multiple parallel orthogonal blocks. Therefore, the detection of data symbols within an orthogonal block only depends on the signals received within this orthogonal block with reduced dimension. Explicitly, we propose a graph theory based orthogonal block identification algorithm, which models the relationship between the received signal and the original information symbols as a bipartite graph, where a depth first search (DFS) algorithm is invoked for partitioning the received signals into orthogonal blocks. For each orthogonal block, the existing detection algorithms can be used. Since the size of orthogonal blocks may be much lower than that of the original received signals, the detection complexity can be significantly reduced. For example, the complexity of the OB based MMSE detector is approximately a factor 4096 lower than that of the traditional MMSE detector for a channel having two paths.