This paper presents a hierarchical framework for detecting local and global anomalies via hierarchical feature representation and Gaussian process regression (GPR) which is fully non-parametric and robust to the noisy training data, and supports sparse features. While most research on anomaly detection has focused more on detecting local anomalies, we are more interested in global anomalies that involve multiple normal events interacting in an unusual manner, such as car accidents. To simultaneously detect local and global anomalies, we cast the extraction of normal interactions from the training videos as a problem of finding the frequent geometric relations of the nearby sparse spatio-temporal interest points (STIPs). A codebook of interaction templates is then constructed and modeled using the GPR, based on which a novel inference method for computing the likelihood of an observed interaction is also developed. Thereafter, these local likelihood scores are integrated into globally consistent anomaly masks, from which anomalies can be succinctly identified. To the best of our knowledge, it is the first time GPR is employed to model the relationship of the nearby STIPs for anomaly detection. Simulations based on four widespread datasets show that the new method outperforms the main state-of-the-art methods with lower computational burden.
Previous studies have shown that the brain network topology correlates with the cognitive function. However, few studies have investigated the relationship between functional brain networks that process sensory inputs and outputs. In this study, we focus on steady-state paradigms using a periodic visual stimulus, which are increasingly being used in both brain-computer interface (BCI) and cognitive neuroscience researches. Using the graph theoretical analysis, we investigated the relationship between the topology of functional networks entrained by periodic stimuli and steady state visually evoked potentials (SSVEP) using two frequencies and eleven subjects. First, the entire functional network (Network 0) of each frequency for each subject was constructed according to the coherence between electrode pairs. Next, Network 0 was divided into three sub-networks, in which the connection strengths were either significantly (positively for Network 1, negatively for Network 3) or non-significantly (Network 2) correlated with the SSVEP responses. Our results revealed that the SSVEP responses were positively correlated to the mean functional connectivity, clustering coefficient, and global and local efficiencies, while these responses were negatively correlated with the characteristic path length of Networks 0, 1 and 2. Furthermore, the strengths of these connections that significantly correlated with the SSVEP (both positively and negatively) were mainly found to be long-range connections between the parietal-occipital and frontal regions. These results indicate that larger SSVEP responses correspond with better functional network topology structures. This study may provide new insights for understanding brain mechanisms when using SSVEPs as frequency tags.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.