Private Smart Space: Cost-Effective ADLs (Activities of Daily Livings) Recognition Based on Superset Transformation

Aging population inspired the market on advanced real time caring for the elder in home setting, accurately recognizing human activities is a challenging task. Activities of daily living are good indicators for behavior recognition. In this paper, we describe a new method to deploy a cost-effective solution which can be run on embedded device as smart router. We use the open dataset, map the raw dataset into a sparse binary matrix, unique by the time line and activity tags. Decision tree algorithm is applied to train the model, in order to achieve the goal that simple comparison work to implement the model and get a quick respond at high accuracy. We evaluate our approach by 3-fold cross validation and achieve a time-slice accuracy of 98.45%.