Signal Model Physical Wave Propagation Channel Model Known Techniques for Channel Prediction Minimum-Energy Bandlimited Prediction Simulation Results Relation to Wiener Filtering Analytic Results

In order to utilize the full capacity of MIMO channels, channel knowledge at the transmitter is necessary. Channel reciprocity can be utilized in time-division duplex (TDD) systems. Prediction is necessary for time-variant channels. How to predict time-variant channels from M noisy channel observations? Lets start with a single antenna frequency flat channel. Thomas Zemen Time-Variant Channel Prediction 3/23 Signal Model Time-variant channel observed at symbol rate RS = 1/TS Symbol duration is much longer than the delay spread of the channel TS TD. Sampled time-variant channel h[m] , h(mTS,0) y [m] = h[m] + z[m] , m ∈ {0,...,M − 1} Channel observed for block length M. Channel coefficients h[m] ∼ CN (0,1). Additive noise z[m] ∼ CN (0, σ2z), σ2z = 1/SNR. Thomas Zemen Time-Variant Channel Prediction 4/23 Physical Wave Propagation (I) vuser receiver scatterer0 scatterer 1 a