Joint estimation of fading channel and data with antenna arrays

In our previous work, we proposed a Kalman filter framework to track fast fading channels for arrays. A limitation of the algorithm was that hard decisions were needed in estimating the channel. We consider three alternative methods for performance improvement, namely, the per-survivor technique, soft decision feedback, and smoothing. It is shown via simulation that the per-survivor technique can approach perfect decision feedback at high SNRs. Soft decision feedback has a decent performance gain over hard decision feedback while incurring slightly increased computation. Smoothing, though complex, has only limited improvement.