WALKING WALKing walking: Action Recognition from Action Echoes

By Qianli Ma, Lifeng Shen, Enhuan Chen, Shuai Tian, Jiabing Wang, Garrison W. Cottrell. 2017

26 citations recorded by the source corpus.

Recognizing human actions represented by 3D trajectories of skeleton joints is a challenging machine learning task. In this paper, the 3D skeleton sequences are regarded as multivariate time series, and their dynamics and multiscale features are efﬁ-ciently learned from action echo states. Speciﬁcal-ly, ﬁrst the skeleton data from the limbs and trunk are projected into ﬁve high dimensional nonlinear spaces, that are randomly generated by ﬁve dynamic, training-free recurrent networks, i.e., the reservoirs of echo state networks (ESNs). In this way, the history of the time series is represented as non-linear echo states of actions. We then use a single multiscale convolutional layer to extract multiscale features from the echo states, and maintain multi-scale temporal invariance by a max-over-time pooling layer. We propose two multi-step fusion strategies to integrate the spatial information over the ﬁve parts of the human physical structure. Finally, we learn the label distribution using softmax. With one training-free recurrent layer and only layer of convolution, our Convolutional Echo State Network (ConvESN) is a very efﬁcient end-to-end model, and achieves state-of-the-art performance on four skeleton benchmark data sets.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

SSNet: Scale Selection Network for Online 3D Action Prediction (2018, background): “Human activities are naturally performed in 3D space, thus 3D skeleton data is suitable for representing human actions [46].”
Global Regularizer and Temporal-Aware Cross-Entropy for Skeleton-Based Early Action Recognition (2018, background): “Given that human actions are performed in the 3D space, 3D skeleton sequences captured by depth cameras provide comprehensive and useful information for action analysis [3].”