Text this: Sparse representations of dynamic scenes for compressive spectral video sensing