A Vector Quantization Approach to LMMSE Channel Estimation for OFDM System

The linear minimum mean square error (LMMSE) channel estimation has better performance than the least square (LS) channel estimation. However, it requires channel statistics which are often difficult to obtain. In addition, the computational complexity of the LMMSE channel estimation is high because of the inverse matrix computation. In order to overcome these implementation issues, a vector quantization approach to LMMSE channel estimation for OFDM systems is proposed in this paper. The LMMSE filter coefficients (LFC) are assumed to be determined by two parameters: root-mean-square delay spread (RDS) and signal-to-noise ratio (SNR). After being calculated by using the LS channel estimation at pilot subcarriers, these parameters are mapped into a representation codevector of the quantization region. So that LFC can be obtained through the pre-calculated look-up table, which stores LFC corresponding to the representation codevector. Optimal vector quantization under the memory constraint is also derived. Finally, the proposed method is applied to channel sounding of the IEEE 802.16e specification and the resulting performance is improved compared to the LS channel estimation.