Engagement plays a key role in improving the cognitive and motor development of children with autism spectrum disorder (ASD). Engaging technologies involving physical activity and interactive stimuli have benefited in improving their engagement and lessening stereotyped behaviors. Especially, content involving both tangible and intangible robot agents can foster engagement in children with developmental disorders. Combining AI systems with these robot agents enables the prediction of engagement and timely interventions, potentially aiding in maintaining high engagement levels. However, the scarcity of data hinders the development of engagement prediction models in practice (i.e., no publicly available dataset for engagement of children with ASD). In this study, we present the Engagnition dataset for engagement recognition of children with ASD (N = 57) utilizing a self-developed serious game, βDefeat the Monsterβ based on enhancing recognition and classification abilities. The dataset consists of physiological and behavioral responses and expertise annotations, based on ternary scales of engagement (ground truth). For technical validations, we report the distribution of engagement, the status of interventions per participant, and the signal-to-noise ratio (SNR) of physiological signals.
Welcome to our GitHub repository for the Developmental Disabilities and Physical Activity Series Game project! π In this project, we have designed and developed a serious game centered around physical activities, specifically tailored for children with developmental disabilities. Through this initiative, we collected valuable data to enhance our understanding of how these games can positively impact the well-being and development of children facing developmental challenges.
Our mission is to create an interactive environment where children with developmental disabilities can participate in physical activities. We believe that gamifying physical exercise can foster engagement and promote essential skills in a playful and supportive manner.
Specifically designed for engagement recognition of kids with Developmental Disabilities (Total participants: 57). Incorporates both physiological and behavioral responses for comprehensive analysis. Contains expert annotations anchored on a ternary scale, serving as the engagement ground truth.
The data for the Engagnition dataset was collected at a physical fitness center specifically designed for children with developmental disabilities.
- E4 Wristband: A versatile biometric sensor capturing a wealth of physiological data. (Accelerometer, GSR, Skin Temperature)
- Unity-Based Game: Participants interacted with a custom-developed game designed in Unity to assess their performance metrics during the sessions. (Performance annotation, Session elapsed times)
- Annotation: Engagement annotation, Gaze fixation annotation
- subjective questionnaires: SUS, NASA-TLX
Data collection encompassed three distinct conditions to provide a multifaceted perspective:
- Baseline: Captures the foundational physiological and behavioral metrics without any specific external stimulus.
- Physical High Demand: Metrics when participants were subjected to tasks or environments demanding high physical engagement.
- Physical Low Demand: Metrics under tasks or conditions that required minimal physical activity or engagement.
Engagement during sessions was classified into three distinct tiers:
- 0: Not engaged at all.
- 1: Moderately engaged.
- 2: Fully engaged.
These engagement classifications were determined in collaboration with experts specializing in developmental disabilities, ensuring a comprehensive and informed categorization.
It's imperative to note that all participants (or their guardians) provided explicit consent for the data to be publicly disclosed and utilized for research purposes.
Email: kimwon30 AT gm.gist.ac.kr, seongminwoo AT gm.gist.ac.kr