Sharing Worlds: Design of a Real-Time Attention Classifier for Robotic Therapy of ASD Children

Abstract

Robotic-assistive therapy has demonstrated very encouraging results for children with Autism. Accurate estimation of the child's pose is essential both for humanrobot interaction and for therapy assessment purposes. Nonintrusive methods are the sole viable option since these children are sensitive to touch. While depth cameras have been used extensively, existing methods face two major limitations: (i) they are usually trained with adult-only data and do not correctly estimate a child's pose, and (ii) they fail in scenarios with a high number of occlusions. Therefore, our goal was to develop a 3D pose estimator for children, by adapting an existing state-of-the-art 3D body modelling method and incorporating a linear regression model to fine-tune one of its inputs, thereby correcting the pose of children's 3D meshes. In controlled settings, our method has an error below 0.3m, which is considered acceptable for this kind of application and lower than current state-of-the-art methods. In real-world settings, the proposed model performs similarly to a Kinect depth camera and manages to successfully estimate the 3D body poses in a much higher number of frames.

A scoping review of artificial intelligence-based robot therapy for children with disabilities

“…The tasks in the therapy sessions required the child to imitate the therapist's and the robot's gestures. These interventions were carried out in an environment that allowed for natural mobility and were recorded using a single camera for later analysis [20]. In the context of the study, Cozmo, a robot, utilizes its display screen, sound, and movements to express various emotional expressions to the participant.…”

“…Specifically, there was a notable improvement in emotional comprehension and a statistically significant enhancement in emotional language abilities [36]. The effectiveness of the therapy is assessed based on changes in the child's ability to engage in joint attention and improvements in communication and social skills, which are often targeted outcomes of joint attention training [20]. Children exposed to the robot decrease their distress and positively change their attitude toward the technological device using the content of the tales (knowledge or emotional) and the different social behaviour of the narrators: static human, static robot, expressive human, and expressive robot [21].…”

“…Inter-rater reliability analysis and sensitivity assessments for the evaluation process, culminate in a comprehensive and quantifiable measurement of joint attention dynamics throughout therapy sessions [48]. This multifaceted approach, which includes manual analysis, automatic classification, tracking, and reliability analysis, enables an in-depth understanding of the impact of therapy techniques on joint attention in children with ASD [20]. Deep learning models such as Yolo v3-tiny and single shot detector (SSD) are utilized for object detection [40].…”