We propose novel multi-fused features for online HAR system.

They are skeleton joint features including joint features DT, DK, M, ⊖ and shape feature HOG-DDS.

The HAR systems recognize human activities from continuous sequences of depth map.

It trains the hidden Markov model (HMM) with the code vectors of the multi-fused features.

It outperforms the state-of-the-art HAR methods in terms of recognition accuracy.