3D Skeletal Human Action Recognition Using a CNN Fusion Model

Figure 7

Sample frames of the NTU RGB + D dataset. The red points in each frame were plotted according to the 3D positions of skeletal joints obtained using the method of [12]. We just select 13 skeletal joints of each pose for convenient viewing. (a) Brushing hair. (b) Dropping. (c) Wearing jacket. (d) Wearing jacket. (e) Patting others’ back. (f) Handshaking.

(a)

(b)

(c)

(d)

(e)

(f)