Abstract-In this paper, the performance of a multi-user multiple-input multiple-output (MIMO) system in time-varying channels is evaluated using measurement data. We consider the multi-user MIMO system using a block diagonalization (BD) scheme and an eigenbeam-space division multiplexing (E-SDM) technique. In an ideal case, the BD scheme eliminates inter-user interference, and the E-SDM technique suppresses inter-stream interference. In actual radio environments, however, channels change over time. This causes interference in the multi-user MIMO system even though the BD scheme and the E-SDM technique are used. To overcome this problem, the authors have developed a simple channel prediction scheme on the basis of a linear extrapolation and have demonstrated its effectiveness by computer simulations assuming the Jakes' model. To verify the performance of the channel prediction scheme in actual environments, we conducted a measurement campaign in indoor environments and measured a large amount of channel data. Using these data, we examined the channel transition and channel tracking with the prediction method. Then we obtained the biterror rate (BER) performance. The prediction technique was shown to track the channel and improve the BER performance almost to that in the ideal time invariant case.