A Behavior Mining Method by Visual Features and Activity Sequences in Institution-based Care

摘要: This paper proposed a behavior mining method through discriminative visual feature extraction from disabled elder caring. Nursing activities for bed-ridden residents were manually annotated video data which was collected and provided by care center. The variation grid sequences are extracted as features the difference blobs of successive image frames. sequential association rules trained based on each nursing activity category. A similarity measure between test segment related is recognition. Through empirical evaluation dataset, approach very promising discrimination features.