Human Depth Sensors-Based Activity Recognition Using Spatiotemporal Features and Hidden Markov Model for Smart Environments

Table 3

Comparison of recognition accuracy of proposed method and state-of-the-art methods using MSRAction3D dataset.

Method	Recognition accuracy
Bag of 3D points [22]	74.7
Shape and motion features [23]	82.1
Eigenjoints [20]	82.3
Spatiotemporal motion variation [24]	84.6
STOP features [25]	84.8
Joints plus body features [26]	85.6
Actionlet ensemble [11]	88.2
Cuboid similar features [27]	89.3
Spatial and temporal part-sets [28]	90.2
Pose-based features [29]	91.5
Proposed method	92.4