Frontal Face Tracking in the Thermal Infrared Imaging for Autism Spectrum Disorder Children

The research and application of therapeutic robotics for children have gained significant attention. This study introduces a system for emotion recognition in Autism Spectrum Disorder (ASD) children that utilizes infrared thermal imaging cameras to capture the frontal facial images. The paper focuses on the development to automate, recognize and track the subjects frontal face using the Shape-Adapted Mean Shift algorithm (SAMShift) applied to thermal images. This system targets the identification of affective states in ASD children and employs non-invasive thermal imaging instead of invasive sensor patches. By integrating tracking and affective state analysis within a single modality, the system addresses the limitations of traditional visual-based systems, which often suffer from invasiveness due to the use of sensor patches, difficulty in automating and tracking facial features, and challenges in accurately identifying affective states. The SAMShift algorithm, implemented in the LabVIEW® vision module environment, is utilized for tracking the frontal face image. Experimental results demonstrate the effectiveness of the proposed method, achieving an average accuracy of 89.20% and an average precision of 95.40%. This research contributes to advancing the field of therapeutic robotics for children with ASD by providing an automated, non-invasive, and accurate approach to emotion recognition.