Design and Analysis of Emotion Elicitation Techniques for ECG-Based Emotion Recognition in Children with Autism Spectrum Disorder (ASD).

Children with autism spectrum disorder (ASD) face difficulties in expressing and recognizing emotions resulting in meltdowns and aggressive situations which is strenuous for both parents and caretakers. A device that would help parents and caretakers to identify the onset of emotional variations would not only aid in preventing undesirable situations but also help them resort to remedial therapies. Towards that, this research work focuses on identifying the positive and negative emotional state of children with ASD using electrocardiogram (ECG) signals and artificial intelligence (AI) techniques. Training AI for emotion recognition requires ECG data to reflect the emotional state. As physiological data cannot be manipulated, emotions need to be induced in children to obtain the emotional ECG data. This research analyses two emotion elicitation protocols namely audio-visual (AV) based emotion elicitation protocol and multimodal emotion elicitation protocol to identify the protocol that can be used to better elicit emotions in children with ASD. In this work, emotional ECG data was acquired from 25 children with ASD using both the protocols. Results indicate that, the ECG data elicited using multimodal emotion elicitation protocol and AV based protocol, categorize emotions at the highest average accuracy of 85.7% and 72.2% respectively. The findings suggest that stronger emotions are elicited using multimodal approach and therefore this approach may be further explored for emotion recognition research and analysis in children with ASD.