ShadowBoard: A Three-Layer Software Framework for Shadow-Robust Hand-Tracking Interactive Whiteboards Using Commodity Webcams

Abstract

Interactive Whiteboards (IWBs) demonstrably improve student engagement and learning outcomes, yet their cost ($1,500–$5,000 per unit) renders them inaccessible to the majority of schools in resource-constrained regions. We present ShadowBoard, a fully implemented, hardware-free interactive whiteboard system operating on any commodity laptop webcam. ShadowBoard addresses the fundamental challenge of visual noise—shadows and reflections induced by unpredictable class-room lighting—through a novel three-layer software architecture. The Shadow Layer converts RGB frames to YUV color space and applies a closed-form adaptive V-channel filter to suppress shadows without distorting hand chrominance. The Landmark Layer feeds denoised frames into MediaPipe Hands to extract a precise 21-point 3-D hand skeleton. The Interactive Layer maps fingertip trajectories to drawing strokes and uses a pinch-distance metric to trigger click and drag gestures. A controlled study with 20 participants across three lighting conditions (80–700 lux) shows ShadowBoard achieves a 68.6% reduction in landmark jitter (6.82 → 2.14 px), 96.2% gesture recognition accuracy, and 24.5 FPS on CPU-only hardware—outperforming raw-RGB and Gaussian Mixture Model baselines on all metrics. An ablation study isolates each layer’s individual contribution. A one-week field deployment at a rural government school in Matara, Sri Lanka (47 students, grades 6–8) confirms ecological validity: field gesture accuracy (94.7%) and jitter (2.31 px) remain within 1.5% and 8% of lab values respectively, while student on-task behaviour rises from 71% to 87% over the deployment week. ShadowBoard requires no auxiliary hardware, making high-quality interactive education accessible to every student regardless of geographic or financial constraints.