In this paper we use a Kinect camera to monitor extensively human movements and evaluate facial expressions in order to implement a smart video surveillance system in a house inhabited with people with Autism Spectrum Disorders. The main problem here is to select the best approach by testing and analyzing several gesture recognition algorithms.Despite the tests are made on people without Autism Spectrum Disorders, its known that people with Autism Spectrum Disorders usually reveal stereotypic motor behaviors and characteristic reactions to social interactions. In most cases the diagnosis techniques are subjective because they rely on the experience of the parents with the children and the time spent with the doctor is very little in order to be made a complete examination to the children's natural behavior [1][2][3].