Low Complexity Block Equalizer for OTFS Based on Expectation Propagation

Abstract

Orthogonal time-frequency space signaling (OTFS) is a novel delay-Doppler (DD) domain modulation scheme that has the potential to exploit two-dimensional time-frequency (TF) diversity. To realize such diversity while retaining low complexity, this letter develops an efficient block equalizer for OFDM-based OTFS via expectation propagation (EP). In particular, the DD-domain channel matrix is decomposed as a function of symbol interval, which has the form of the superimposed diagonal-times-doubly circulant matrices. We exploit such decomposition to facilitate the computation of the moment-matching step for channel factor update, reducing the overall equalizer complexity order from cubic to log-linear while achieving almost the same bit-error-rate (BER) performance as compared to the exact computing case.