Information and communication technologies have contributed to the social and cognitive stimulation of children with autism spectrum disorder (ASD). This seems to be partly because children with ASD are more comfortable with predictive and repetitive behaviours [1], which can be implemented through algorithms. In particular, artificial intelligence algorithms are used in various fields, from language or gesture recognition to image classification [2]. These functions are convenient for a system to interact with a human being. Furthermore, robots seem to prove useful with autistic children because of simpler face expressions than those of human beings [3]. Robots can perform gesture imitation learning [4]. For a gesture to be recognized, the human body must first be correctly represented. Joint angles and joint positions can be used to represent human motion in spaces more suitable than the Euclidian one.