Comment on "The Kalman filter and nonlinear estimates of multivariate normal processes"