The present invention is a system for detecting, tracking and recognizing human faces in a visual prosthesis. In a visual prosthesis, the input camera is always higher resolution than the electrode array providing percepts to the subject. It is advantageous to detect, track and recognize human faces. Then information can be provided to the subject by highlighting the face in the visual scene, providing auditor or vibratory notice that a human face is in the visual scene, looking up the face in a database to state the name of the person in the visual scene, otherwise communication id like providing a custom vibratory pattern corresponding to known individuals (like custom ring tones associated with caller ID) or automatically zooming in on a face to aid the subject in identifying the face.