Efficient convolutional hierarchical autoencoder for human motion prediction

• DOI: 10.1007/s00371-019-01692-9
• Publication Date: 2019-05-11T12:41:22Z
• Authors: Yanran Li; Zhao Wang; Xiaosong Yang; Meili Wang; Sebastian Iulian ...; Ehtzaz Chaudhry; Jianjun Zhang
• Keywords: Autoencoder; Benchmark (surveying)

Human motion prediction is a challenging problem due to the complicated human body constraints and high-dimensional dynamics. Recent deep learning approaches adopt RNN, CNN or fully connected networks learn features which do not exploit hierarchical structure of anatomy. To address this problem, we propose convolutional autoencoder model for with novel encoder incorporates 1D layers topology. The new network more efficient compared existing models respect size speed. We train generic on Human3.6M CMU benchmark conduct extensive experiments. qualitative quantitative results show that our outperforms state-of-the-art methods in both short-term long-term prediction.