High quality facial expression recognition in video streams using shape related information only.