Hierarchical Aligned Cluster Analysis for Temporal Clustering of Human Motion

Abstract

Abstract-Temporal segmentation of human motion into plausible motion primitives is central to understanding and building computational models of human motion. Several issues contribute to the challenge of discovering motion primitives: the exponential nature of all possible movement combinations, the variability in the temporal scale of human actions, and the complexity of representing articulated motion. We pose the problem of learning motion primitives as a temporal clustering one, and derive an unsupervised hierarchical bottom-up framework called hierarchical aligned cluster analysis (HACA). HACA finds a partition of a given multidimensional time series into m disjoint segments such that each segment belongs to one of k clusters. HACA combines kernel k-means with the generalized dynamic time alignment kernel to cluster time series data. Moreover, it provides a natural framework to find a low-dimensional embedding for the time series. HACA is efficiently optimized with a coordinate descent strategy and dynamic programming. Experimental results on motion capture and video data demonstrate the effectiveness of HACA for segmenting complex motions and as a visualization tool. We also compare the performance of HACA to state-of-the-art algorithms for temporal clustering on data of a honey bee dance. The HACA code is available online.

A Temporal Clustering Algorithm for Achieving the Trade-off between the User Experience and the Equipment Economy in the Context of IoT

“…The authors make a rather complete survey proposing a categorization of the main components that characterize the task of grouping a time series. Despite the completeness of this and other studies [5,6], the focus of the authors often was on the efficiency and complexity of the approaches in the context of big data and cloud computing. Our context is different because we seek to solve a time series prediction problem with severe processing and storage limitations.…”

A review of computational approaches for evaluation of rehabilitation exercises

“…These methods determine segments based on zero crossings for the velocity [61], [62] or acceleration [149] of joint trajectories. Distance functions, such as Euclidean distance [150], Mahalanobis distance [151], and DTW distance [152] have also been used for this purpose, where segments are extracted at the points having the value of the distance function greater than a pre-selected threshold. Lee et al [153] introduced a deep learning-based approach for segmentation of time-series, in which an autoencoder network extracted representative features from input data, and the peaks in a distance function calculated from the features were selected as breakpoints for segmentation purpose.…”

Temporal clustering of surgical activities in robot-assisted surgery

“…In particular, we apply models developed for human activity recognition [17, 18]. We explore our models using data from clinically relevant tasks performed on porcine models in a training environment.…”