OTFS Detection Based on Gaussian Mixture Distribution: A Generalized Message Passing Approach

The Orthogonal Time Frequency Space (OTFS) modulation has been identified as a promising modulation scheme for wireless communications with high mobility. As for data detection of OTFS symbols, most existing detectors rely on the sophisticated message passing (MP) algorithm and its variants. In particular, the classic sum-product (SP)-MP algorithm can achieve a good detection performance with high complexity, while the Gaussian approximation-based (GA)-MP detector reduces the complexity significantly at the cost of performance degradation. In order to achieve a more flexible detection framework, we propose to adopt the Gaussian mixture distribution (GMD) for representing messages on the factor graph. The proposed scheme can behave as either the classic SP-MP algorithm or the GA-MP algorithm by varying the number of adopted Gaussian components. The simulation results show the proposed scheme can strike a good tradeoff between the error performance and complexity.