Blind and semi-blind channel estimation with fast convergence for MIMO-OFDM systems

• We develop the block matrix scheme for subspace channel estimation of the CP-OFDM system.
• We establish the identifiability conditions for the blind subspace channel estimation.
• We present the semi-blind subspace channel estimation.
• We analyze the computational complexity of the proposed blind/semi-blind channel estimation
• The simulation results for both the time-invariant and the time-varying scenarios are performed.

In this paper, the blind subspace channel estimation using the block matrix scheme is proposed for multiple-input multiple-output (MIMO) orthogonal frequency division multiplexing (OFDM) systems. Based on the Toeplitz structure, the block matrix scheme collects a group of the received OFDM symbols into a vector, and then partitions it into a set of equivalent symbols. The number of equivalent symbols is about N times of OFDM symbols, where N is the size of FFT operation. With those equivalent symbols, the proposed blind subspace channel estimation can converge within a small amount of OFDM symbols. The identifiability of the proposed channel estimation is examined that the channel matrix is determined up to an ambiguity matrix. Besides, the semi-blind channel estimation is also investigated by combining few pilot sequences with the subspace method. Simulation results show that the proposed channel estimators perform very well even in a time-varying scenario. Especially the semi-blind methods achieve almost the same BERs as those by true channels.