Spike train driven dynamical models for human actions

Abstract

We investigate dynamical models of human motion that can support both synthesis and analysis tasks. Unlike coarser discriminative models that work well when action classes are nicely separated, we seek models that have finescale representational power and can therefore model subtle differences in the way an action is performed. To this end, we model an observed action as an (unknown) linear time-invariant dynamical model of relatively small order, driven by a sparse bounded input signal.Our motivating intuition is that the time-invariant dynamics will capture the unchanging physical characteristics of an actor, while the inputs used to excite the system will correspond to a causal signature of the action being performed. We show that our model has sufficient representational power to closely approximate large classes of non-stationary actions with significantly reduced complexity. We also show that temporal statistics of the inferred input sequences can be compared in order to recognize actions and detect transitions between them.

Substructure and boundary modeling for continuous action recognition

“…Quaternion feature is derived from the raw MoCap data as our observation for inference. Table 2 lists the continuous action recognition results, in comparison with the same set of benchmark techniques as in the first experiment, as well as [29,30]. Similarly, results from this experiment demonstrated the superior performance of our method.…”

Metric Learning from Poses for Temporal Clustering of Human Motion

“…They show that ACA can accurately find different behaviors in sequences of mocap data. However, [12] [22] are limited by having to manually set the number of clusters (actions) k. In [13], this limitation is tackled by using a spike-train driven dynamical model that can detect motion transitions and clusters them into different behaviors, without having to manually set the number of clusters k. As far as video data is concerned, approaches such as [9][11] have proposed variants and extensions of hierarchical Dirichlet processes (HDP) [17] in order to find activities using optical flow features mainly. In [7], HDPs are used as a prior for HMM parameters in order to cluster time series data into distinct behaviors.…”

Extraction of Information from Inertial Sensors to Aid Health Assessment

“…The dynamical model used for estimation of stimuli from the inertial data has been inspired from Raptis' work [53]. The model, proposed in [53], falls into the class of linear dynamical models, where the task of motion modeling has been posed as a system identification problem under bounded energy and sparsity constraints [54].…”

“…The model, proposed in [53], falls into the class of linear dynamical models, where the task of motion modeling has been posed as a system identification problem under bounded energy and sparsity constraints [54]. The motivations to employ this model are two-fold.…”

“…This forces variables such as the amplitude of an action into the system matrices, thus resulting in inputs that are more comparable across individuals. An expectation -maximization (EM) algorithm [53] was applied to estimate the , , , and matrices under the assumption of system linearity and time invariance.…”