Recognition of human physical activity based on a novel hierarchical weighted classification scheme

The automatic recognition of postures, movements and physical exercises has being recently applied to several healthcare related fields, with a special interest in chronic disease management and prevention. In this work we describe a complete method to define an accurate activity recognition system, stressing on the classification stage. As binary classifiers can be, in general, considered more efficient than direct multiclass classifiers, and looking for an appropriate multiclass extension schema, a hierarchical weighted classification model with a special application for multi-sensed problems is presented. Remarkable accuracy results are obtained for a particular activity recognition problem in contrast to a traditional multiclass majority voting algorithm.