In this talk, we will introduce a reduced dimensionality representation for the intrinsic dynamic features of an underlying process, which is measured from different channels through different sensors. In order to eliminate the effects of nuisance and sensor-specific parameters, we consider a random walk on the data points, which is restrained to hop between different views. Based on this cross-view model, we define a multiview diffusion distance, which is insensitive to small structural changes in the data and leads to reduced dimensionality representations not only for each view but also for the intrinsic dynamic. The application of the multiview diffusion map on data clustering will also be discussed.