Artificial Vision Algorithm for Behavior Recognition in Children with ADHD in a Smart Home Environment

Artificial vision has made a great advance in the recognition of visual patterns that are not perceptible by humans or that are biased in their interpretation. Among its applications, artificial vision or computer vision has served in the support of people with some kind of disability. In this work, an image classification algorithm is developed to complement a pervasive therapy support system for children with Attention Deficit Hyperactivity Disorder (ADHD) during the development of their homework. For this purpose, a camera is adapted within a smart environment made up of Smart objects and a robotic assistant. In the system, a convolutional neural network (CNN) is implemented for the classification of the child’s status (doing or not doing his/her homework). An experiment of this implementation is carried out in which the results of the environment without the camera are compared with the results obtained by using the camera and the implemented CNN. The latter results are also compared with the information collected through observation by the therapist during the session. The results show that what the camera identifies as the child not doing homework matches what the smart objects identify as distractions and pauses at 82.70% and what the therapist identifies as distractions and pauses at 98.21%. This approach will help the smart home environment have new and more accurate data to process and make better decisions, just like a therapist would do.