If it helps, footprint data are available here.
The hypersensitivity track data contains regions of DNaseI hypersensitivity (DHSs) called through a process developed in the Stamatoyannopolis lab. The Nature Methods paper by Sabo et al. describes this in more detail.
The raw signal are windowed regions of chromatin accessibility across a genome, with the density of the 5' end of DNase cut fragments within each 150 bp window, every 20 bp.
Peaks and hotspots are called from these regions.
Within nearly all DHSs, there is often one or more footprints found. This footprint is a relatively shorter region of variable length (6-40 bp, I think) where DNase does not cleave, because of bound proteins or protein complexes (or related phenomena, like lack of methylation, which allows proteins to bind DNA), like transcription factors or transcription initiation machinery.
See the Nature ENCODE paper by Neph et al., also by the same lab, which explains this in more detail.
If you're doing TF prediction, the footprints will likely be of more use to you. Shane Neph's de novo motif discovery algorithm used these footprints to discover binding sites for 683 motifs. 394 of them matched entries in non-redundant and reduced known and experimental transcription factor databases (TRANSFAC. JASPAR Core, UniPROBE, and some data sets from the Kellis lab). The remaining 289 were found to be novel, frequented millions of footprints, and showed similarities with known TFs.