WLA4ND: A Wearable Dataset of Learning Activities for Young Adults with Neurodiversity to Provide Support in Education

Abstract

Data-driven assistive wearable technologies are promising to support young adults with neurodiversity in inclusive education. Existing datasets focus on wearable sensor data of various activities from neurotypical people. However, no dataset exists including learning-related activity data from individuals with neurodiversity. The contributions of this paper include (1) WLA4ND, a dataset of learning activities performed by eight young adults with neurodiversity collected from smartwatch sensors. The activities are common learning tasks, including reading, writing, typing, answering follow-up questions, and off-task. (2) Evaluation of classification on WLA4ND with five activity recognition models, including conventional and deep learning methods. The Convolutional Recurrent Neural Network (CRNN) model achieved a balanced accuracy of 92.2% for user-dependent evaluations, while Federated Multi-Task Hierarchical Attention Model (FATHOM) achieved 91.8% for user-independent evaluations. This evaluation demonstrates that existing activity recognition technologies can be applied to neurodiverse populations. Also, WLA4ND can be used by researchers as a complement for activity recognition, automatic labeling, and next-generation assistive wearable applications.