Face recognition is certainly the most natural approach to person recognition in smart home environments. The extreme variability of faces in such applications, due to continuous changes in terms of pose, illumination and subject appearance (hairstyle, make-up, etc.), requires very robust and fast algorithms to be developed. Moreover the variations of the subject's face cannot usually be adequately encoded in the initial user template, typically created starting from a few example images, thus making necessary to continuously update the templates on the basis on new inputs. After a review of the state of art of video-based face recognition approaches, suitable for home environments, a semi-supervised video-based template updating approach introduced by the authors is presented.