Algorithm 2. The Pseudo-code of Activity Recognition Using the Hierarchical Framework

Inputs:

(1) a test sample x

(2) a set of classifiers and feature subsets cls_fs

Output:

(1) the activity label of x

1. set the root node as current node nd; //initialization

2. obtain the number of children of nd, and note it as |↓(nd)|;

3. project x over Snd, and use clsnd to get the next-level label PL and corresponding next-level node pnd, in which 1 ≤ pnd ≤ |↓(nd)| and PL corresponds to the maximal probability output of clsnd

4. if is leaf_node(k) // conditional statement

(4.1) if TRUE do   return PL as the predicted activity label;

(4.2) if FALSE do   set node pnd as current node, and go to step 2;