An adaptive MIMO channel estimation based on Kalman filtering is proposed in this paper. The process of Excess Error Energy varying with iteration number is analyzed. The numerical simulation of Kalman channel estimation and Maximum Likelihood (ML) channel estimation based on orthogonal training sequences applying in saturated and subsaturated system is presented. The simulation results show that the performance of Kalman channel estimation proposed herein is closed to that of ML channel estimation and Kalman estination works better in saturated MIMO system.