View- and Rate-Invariant Human Action Recognition