Soft Information Aided Diagonal Kalman Filter for Joint Channel Estimation and Detection in Massive MIMO Systems

This paper proposes a soft-information-aided diagonal Kalman filter (SIA-DKF) for joint channel estimation and symbol detection (JED) in massive multipleinput multiple-output (MIMO) systems. Specifically, building upon the Kalman filter (KF) framework for JED, the proposed SIA-DKF incorporates the symbol means and variances obtained from variational inference (VI) into the KF, improving its ability to track channels and mitigating mismatches in the update step. Furthermore, SIA-DKF not only exploits favorable propagation in massive MIMO to simplify the VI for symbol detection, but also employs a diagonal approximation of the covariance matrix at each channel update step, thereby significantly reducing the computational complexity. Simulations confirm that SIA-DKF improves the bit error rate (BER) with lower complexity.