Two-Dimensional DPS-BEM Based Channel Estimation for MIMO OTFS Systems

In this paper, a novel doubly selective channel estimation (CE) scheme is proposed for multiple-input multiple-output (MIMO) orthogonal time frequency space (OTFS) systems. Using the two-dimensional (2D) discrete prolate spheroidal basis expansion model (DPS-BEM), a new analytical representation of the MIMO channel is derived in the delay-Doppler domain. This results in a significant reduction of the number of unknown parameters required for CE compared to methods that directly estimate the channel. Moreover, inspired by the pilot structure in recent works, a new low-overhead pilot scheme is introduced which is capable of effectively capturing the channel's temporal variations and enhancing the estimation performance of the proposed method. The accuracy of the proposed 2D DPS-BEM CE method is evaluated by deriving the theoretical Cramer-Rao lower bound, which shows a low estimation error. Also, the proposed method's computational complexity is calculated and compared with other recent techniques, demonstrating lower complexity. Simulation results further validate that the proposed method outperforms the existing CE approaches in terms of normalized mean squared error and bit error rate, verifying the effectiveness of the proposed 2D DPS-BEM MIMO channel representation.