Institution profile

Toronto Rehabilitation Institute

Academic institutionnorthamerica · ca
Official website
Research library1linked papers
Opportunities0open roles
Selected work

Representative Papers

OPEN: A Benchmark Dataset and Baseline for Older Adult Patient Engagement Recognition in Virtual Rehabilitation Learning Environments

Jul 23, 2025

Reliable measurement of engagement in virtual rehabilitation groups for older adults remains an open challenge. Method: This study introduces OPEN, the first multimodal engagement benchmark dataset specifically designed for long-term virtual rehabilitation training in elderly populations. Distinct from prior work, OPEN uniquely integrates behavioral features—including facial expressions, hand gestures, and body joint keypoints—while incorporating context-aware labels and multi-granularity temporal annotations to explicitly model situational dependencies and cross-session longitudinal engagement patterns. Built upon 35 hours of real-world clinical data, the dataset supports engagement recognition via machine learning and deep learning models augmented with affective and behavioral analysis. Results: The proposed framework achieves 81% accuracy in engagement classification. OPEN and its associated modeling paradigm provide a scalable foundation and a novel methodological framework for personalized, intelligent interventions in remote healthcare settings.

0 citationsRead paper
Recent publications

Latest Papers

OPEN: A Benchmark Dataset and Baseline for Older Adult Patient Engagement Recognition in Virtual Rehabilitation Learning Environments

Jul 23, 2025

Reliable measurement of engagement in virtual rehabilitation groups for older adults remains an open challenge. Method: This study introduces OPEN, the first multimodal engagement benchmark dataset specifically designed for long-term virtual rehabilitation training in elderly populations. Distinct from prior work, OPEN uniquely integrates behavioral features—including facial expressions, hand gestures, and body joint keypoints—while incorporating context-aware labels and multi-granularity temporal annotations to explicitly model situational dependencies and cross-session longitudinal engagement patterns. Built upon 35 hours of real-world clinical data, the dataset supports engagement recognition via machine learning and deep learning models augmented with affective and behavioral analysis. Results: The proposed framework achieves 81% accuracy in engagement classification. OPEN and its associated modeling paradigm provide a scalable foundation and a novel methodological framework for personalized, intelligent interventions in remote healthcare settings.

0 citationsRead paper