Abstract
The conventional minimum variance distortionless response (MVDR) algorithm proposed by Capon for blind beamforming and direction-of-arrival (DOA) estimation, and the multiple signals classification (MUSIC) algorithm proposed by Schmidt only for DOA estimation have been widely used in array signal processing areas. A fun-damental assumption made by the two algorithms is that all the unknown source sig-nals are wide-sense stationary and mutually statistically independent. Beamforming has been considered in the uplink of a multiuser orthogonal frequency division multi-plexing (OFDM) system, while this assumption is only valid for the pre-FFT beam-forming structure. However, the resultant beamforming performance of either MVDR algorithm or MUSIC algorithm is limited due to lack of path diversity. By properly formulating a discrete-time multi-input multi-output signal model for the post-FFT beamforming structure of the uplink of a multiuser OFDM system, MVDR and MUSIC algorithms can be shown to be applicable if ensemble average for estimating data correlation matrices is replaced by subcarrier average (i.e., the above fundamental assumption can be relaxed). Then a semi-blind channel estimation algorithm by subcarrier averaging is proposed which needs only one OFDM block with few pilot subcarriers. The proposed algorithm basically consists of DOA estima-tion through using MUSIC algorithm, source extraction through using MVDR algo-rithm, estimation of time delay and fading gain of each DOA, and their classification (associated with each user) as well as channel recovery of each user. Finally, some simulation results are presented to support the effectiveness of the proposed semi-blind channel estimation algorithm.