SINR Estimation in Random Beamforming with Noisy MIMO Channel Measurements

An improved SINR metric is proposed for the random beamforming scheme introduced by Sharif and Hassibi, when the channel observation used to compute the SINR is known to be noisy or outdated. The effect of noise on the MIMO channel estimate is accounted for using results on the perturbation of the eigenspaces of Hermitian matrices. The new metric, designed as a conservative estimate of the real SINR, is based on expectations of bounds on the signal and interference power. It is shown through simulations that it can noticeably reduce the outage probability, in the realistic setting of a 4 times 2 antennas system, at the cost of a minor reduction of the achievable sum-rate.