Room-Aware Transformer for Activities of Daily Living Recognition in Adults with Mild Neurocognitive Disorder Using Ambient Sensors: Development and Evaluation Study

A room-aware Transformer for ambient-sensor ADL recognition in adults with mild neurocognitive disorder matched or exceeded global and LSTM performance in 9 of 15 home-window comparisons.

• Why it matters: Real-world ADL recognition is hindered by differing home layouts, sensor setups, routines, and imbalanced activity classes, limiting reliable longitudinal assessment of everyday functioning.
• What they did: Using motion, temperature, illuminance, humidity, and door-contact data aligned to activity logs from five homes (92,765 minute-level observations; 67 home-days), models were trained per home with 10-, 30-, and 60-minute windows and evaluated on held-out test days.
• The result: The room-aware Transformer’s accuracy ranged 0.27–0.57 with micro-AUC 0.57–0.82 and macro-AUC 0.54–0.78, with clearest gains in Homes C and E at 10 minutes and Home A at 60 minutes, while Sleep/Rest was most consistently recognized.

AI summary of the abstract · may contain errors · not medical advice · read the paper.