A Computational Model of “Active Vision” for Visual Search in Human–Computer Interaction