Spatio-temporal Location Research Articles

VideoLSTM convolves, attends and flows for action recognition

We present VideoLSTM for end-to-end sequence learning of actions in video. Rather than adapting the video to the peculiarities of established recurrent or convolutional architectures, we adapt the architecture to fit the requirements of the video medium. Starting from the soft-Attention LSTM, VideoLSTM makes three novel contributions. First, video has a spatial layout. To exploit the spatial correlation we hardwire convolutions in the soft-Attention LSTM architecture. Second, motion not only informs us about the action content, but also guides better the attention towards the relevant spatio-temporal locations. We introduce motion-based attention. And finally, we demonstrate how the attention from VideoLSTM can be exploited for action localization by relying on the action class label and temporal attention smoothing. Experiments on UCF101, HMDB51 and THUMOS13 reveal the benefit of the video-specific adaptations of VideoLSTM in isolation as well as when integrated in a combined architecture. It compares favorably against other LSTM architectures for action classification and especially action localization.

A Trust-Based Coordination System for Participatory Sensing Applications

Participatory sensing (PS) has gained significant attention as a crowdsourcing methodology that allows ordinary citizens (non-expert contributors) to collect data using low-cost mobile devices. In particular, it has been useful in the collection of environmental data. However, current PS applications suffer from two problems. First, they do not coordinate the measurements taken by their users, which is required to maximise system efficiency. Second, they are vulnerable to malicious behaviour. In this context, we propose a novel algorithm that simultaneously addresses both of these problems. Specifically, we use heteroskedastic Gaussian Processes to incorporate users' trustworthiness into a Bayesian spatio-temporal regression model. The model is trained with measurements taken by participants, thus it is able to estimate the value of the phenomenon at any spatio-temporal location of interest and also learn the level of trustworthiness of each user. Given this model, the coordination system is able to make informed decisions concerning when, where and who should take measurements over a period of time. We empirically evaluate our algorithm on a real-world human mobility and air quality dataset, where malicious behaviour is synthetically produced, and show that our algorithm outperforms the current state of the art by up to 60.4% in terms of RMSE while having a reasonable runtime.

Protease cleavage site fingerprinting by label-free in-gel degradomics reveals pH-dependent specificity switch of legumain.

Determination of protease specificity is of crucial importance for understanding protease function. We have developed the first gel-based label-free proteomic approach (DIPPS-direct in-gel profiling of protease specificity) that enables quick and reliable determination of protease cleavage specificities under large variety of experimental conditions. The methodology is based on in-gel digestion of the gel-separated proteome with the studied protease, enrichment of cleaved peptides by gel extraction, and subsequent mass spectrometry analysis combined with a length-limited unspecific database search. We applied the methodology to profile ten proteases ranging from highly specific (trypsin, endoproteinase GluC, caspase-7, and legumain) to broadly specific (matrix-metalloproteinase-3, thermolysin, and cathepsins K, L, S, and V). Using DIPPS, we were able to perform specificity profiling of thermolysin at its optimal temperature of 75°C, which confirmed the applicability of the method to extreme experimental conditions. Moreover, DIPPS enabled the first global specificity profiling of legumain at pH as low as 4.0, which revealed a pH-dependent change in the specificity of this protease, further supporting its broad applicability.

An electromyographic study on the sequential recruitment of bilateral sternocleidomastoid and masseter muscle activity during gum chewing.

Mandibular functions are associated with electromyographic activity of the jaw muscles and also the sternocleidomastoid muscle (SCM). The precise spatiotemporal relation of SCM and masticatory muscles activities during chewing is worthy of investigation. To analyse the sequential recruitment of SCM and masseter activities during chewing as indicated by the spatiotemporal locations of their activity peaks. Jaw movements and bilateral surface electromyographic activity of SCM and masseter were recorded during gum chewing in 20 healthy subjects. The timing order was decided by comparing the length of time from the time when the opening started to the time when the surface electromyographic activity reached its peak value. Spatial order was analysed by locating the peak electromyographic activity onto a standard chewing cycle which was created based on 15 unilateral chewing cycles. Paired t-test, one-way ANOVA and Student-Newman-Keuls post-test were used for comparisons. Although the Time to Peak for the balancing side SCM appeared shorter than for the other three tested muscles, most often it did not reach a level of significance. However, the location of the balancing side SCM's peak activity was further from the terminal chewing position (TCP) than the working side SCM and bilateral masseters (P<0·05). The balancing side SCM activity reached its peak significantly further away from TCP than the other three tested muscles during chewing. Further studies with spatiotemporal variables included should be helpful to understand the roles of the head, neck and jaw muscles in orofacial and cervical dysfunctional problems.

Modulation of chaperone-like and membranolytic activities of major horse seminal plasma protein HSP-1/2 by L-carnitine.

The major protein of horse seminal plasma, HSP-1/2, exhibits membranolytic and chaperone-like activities and plays a crucial role in regulating sperm capacitation. L-Carnitine is a small polar molecule present in high concentrations in mammalian seminal plasma. The present results demonstrate that L-carnitine binds to HSP-1/2 and increases its thermal stability, enhances cooperativity of its chemical unfolding and decreases both chaperone-like and membranolytic activities of this protein. The HSP-1/2-L-carnitine complex exhibits anti-oxidative behaviour by inhibiting the production of hydroxyl radicals, suggesting that it can protect other constituents of seminal plasma from damage by hydroxyl radicals. As HSP-1/2 and L-carnitine share the same spatiotemporal location in the horse reproductive tract, this interaction is physiologically significant and may prevent premature interaction of HSP-1/2 with sperm, which in turn regulates the sperm capacitation.

Re-Thinking Gareth Evans’ Approach to Indexical Sense and the Problem of Tracking Thoughts

In “Understanding Demonstratives”, Gareth Evans bites the bullet regarding Rip van Winkle cases in cognitive dynamics: the fact that Rip sleeps for twenty years and completely loses track of time means he is unable to retain his original belief that “Today is a fine day”. In this paper, the author argues that Evans need not bite this bullet because there are resources in his account of the cognitive dynamics involved in belief retention developed in The Varieties of Reference to successfully confront the challenge posed by the Rip van Winkle case. In particular, when we combine the two central elements of Evans’s cognitive dynamics – the skill of keeping track of one’s spatio-temporal location in addition to memory – it is possible to arrive at the conclusion that it is indeed possible for Rip to retain and re-express his original belief.

The geometry of space-time prisms with uncertain anchors

ABSTRACTSpace-time prisms envelop all spatio-temporal locations that moving objects may have visited between two of their known spatio-temporal locations, given a bound on their travel speed. In this context, the known locations are often the result of observations or measurements, and they are called ‘anchor points’. The classic space-time prism, in isotropic two-dimensional space, as well as in transportation networks, assumes that the measurements of these anchor points are exact. Whereas, in many applications, we can assume that time can be measured fairly precisely, this assumption is unrealistic for the spatial components of measured locations (we think of Global Positioning System (GPS) errors, for instance). In this paper, we extend the classical prism from anchor points to circular ‘anchor regions’ that capture the uncertainty or error on their measurement. We define the notion of a space-time prism with uncertain anchor points, called uncertain prism, for short. We study the geometry of uncertain prisms in an arbitrary metric space to make this concept as widely applicable as possible. We also focus on the rims of uncertain space-time prisms, which demarcate the area that a moving object can have visited between two anchor regions (given some local speed limitations).

Enhanced skeleton visualization for view invariant human action recognition

Human action recognition based on skeletons has wide applications in human–computer interaction and intelligent surveillance. However, view variations and noisy data bring challenges to this task. What’s more, it remains a problem to effectively represent spatio-temporal skeleton sequences. To solve these problems in one goal, this work presents an enhanced skeleton visualization method for view invariant human action recognition. Our method consists of three stages. First, a sequence-based view invariant transform is developed to eliminate the effect of view variations on spatio-temporal locations of skeleton joints. Second, the transformed skeletons are visualized as a series of color images, which implicitly encode the spatio-temporal information of skeleton joints. Furthermore, visual and motion enhancement methods are applied on color images to enhance their local patterns. Third, a convolutional neural networks-based model is adopted to extract robust and discriminative features from color images. The final action class scores are generated by decision level fusion of deep features. Extensive experiments on four challenging datasets consistently demonstrate the superiority of our method.

Patterns, Information, and Causation

This paper articulates an account of causation as a collection of information-theoretic relationships between patterns instantiated in the causal nexus. I draw on Dennett’s account of real patterns to characterize potential causal relata as patterns with specific identification criteria and noise tolerance levels, and actual causal relata as those patterns instantiated at some spatiotemporal location in the rich causal nexus as originally developed by Salmon. I develop a representation framework using phase space to precisely characterize causal relata, including their degree(s) of counterfactual robustness, causal profiles, causal connectivity, and privileged grain size. By doing so, I show how the philosophical notion of causation can be rendered in a format that is amenable for direct application of mathematical techniques from information theory such that the resulting informational measures are causal informational measures. This account provides a metaphysics of causation that supports interventionist semantics and causal modeling and discovery techniques.

Locative Inversion and Stage Topics: A Cross-Linguistic Study

This paper seeks to identify the conditions which a clause-initial X phrase (XP) must satisfy to license locative inversion. Three SVO (subject-verb-object) languages with distinct degrees of word order flexibility are considered: English, French and European Portuguese (EP). We show that these languages only permit locative inversion with preposed XPs which are stage topics (sTOPs), i.e. presupposed XPs that define a (spatio-temporal) location. We further show that being a sTOP is both a necessary and sufficient condition for an XP to license this type of inversion. In addition to the structures typically classified as locative inversion, cases of “absolute inversion” in French, “free” inversion in sentence-focus contexts in EP, and inversion structures with preposed participial and adjectival phrases in English and French are argued to be licensed by (c)overt sTOPs and, thus, constitute forms of (covert) locative inversion. We conclude that there is a strong correlation between sTOPs and subject-verb inversion cross-linguistically.

Optimal Retail Location: Empirical Methodology and Application to Practice

We empirically study the spatio-temporal location problem motivated by an online retailer that uses the Buy-Online-Pick-Up-In-Store fulfillment method. Customers pick up their orders from trucks parked at specific locations during specific times, and the retailer’s problem is to determine where and when these pick-ups occur. We combine demographic and economic data, business location data, and the retailer’s historical sales and operations data to predict demand at potential locations. We introduce a novel two step procedure that combines machine learning and conometric techniques. First, we use a random forests algorithm to predict demand when a particular location operated in isolation. Then, we use a fixed effects regression to estimate spatial and temporal cannibalization effects that cannot be captured in the first step. Based on the predicted demand, we develop heuristics to improve the pick-up location configuration and schedule. We estimate a 36% increase in revenue from the improved location configuration and schedule.

Spatiotemporal social (STS) data model: correlating social networks and spatiotemporal data

A location-based social network is a network representation of social relations among actors, which not only allow them to connect to other users/friends but also they can share and access their physical locations. Here, the physical location consists of the instant location of an individual at a given timestamp and the location history that an individual has accumulated in a certain period. This paper aimed to capture this spatiotemporal social network (STS) data of location-based social networks and model it. In this paper, we propose a STS data model which captures both non-spatial and spatial properties of moving users, connected on social network. In our model, we define data types and operations that make querying spatiotemporal social network data easy and efficient. We extend spatiotemporal data model for moving objects proposed in Ferreira et al. (Trans GIS 18(2):253–269, 2014) for social networks. The data model infers individual’s location history and helps in querying social network users for their spatiotemporal locations, social links, influences, their common interests, behavior, activities, etc. We show the some results of applying our data model on a spatiotemporal dataset (GeoLife) and two large real-life spatiotemporal social network datasets (Gowalla, Brightkite) collected over a period of two years. We apply the proposed model to determine interesting locations in the city and correlate the impact of social network relationships on the spatiotemporal behavior of the users.

The utility of EMR address histories for assessing neighborhood exposures

PurposeElectronic medical records (EMRs) include residential address histories, which may alleviate exposure misclassification caused by exclusion of patient spatiotemporal location. EMR data are increasingly available but rarely leveraged as a measure of cumulative environmental exposure, in part due to limited understanding of the validity of EMR-derived address histories. MethodsWe compared EMR address histories to self-reported histories among 100 patients of a safety-net health care system completing a telephone survey. We assessed agreement and compared seven neighborhood-level environmental exposures as assessed using both data sources. ResultsWhile 17.1% of respondents did not live at the most recent EMR-derived address during the survey, nearly all (98%) lived there at some point. For respondents with more than one EMR-derived address (N = 64), 87.5% had once lived at the previous EMR address. Of these, 30.4% lived at 1 or more additional residences between the two most recent EMR address. For all measures, neighborhood-level environmental exposures did not differ when using EMR-derived versus self-report addresses. ConclusionsMore recent EMR-derived addresses are more accurate, and differences compared to self-reported addresses in neighborhood-level exposures are negligible. EMR-derived address histories are incomplete and likely suffer from collection bias; future research should further assess their validity and reliability.

The effect of population-scale human activity patterns on exposure to air pollution

Introduction. Exposure is often estimated only at the patient home address, inherently suggesting that subjects stay near their residence throughout the day. Methods. We studied exposure to nitrogen oxides of large groups of individuals by anonymously tracking their location during the day. Differences in exposure at the home address and while accounting for the time spend at work/school were studied. Geocoded residence and school addresses of 1.5 million schoolchildren age 6-18 and of the workplace of 400,000 adults were obtained from the Israeli Central Bureau of Statistics. Air pollution levels for 2008 were obtained using two models: a stationary nationwide land use regression model for NOx concentrations (250 m grid resolution), and a dynamic optimization dispersion model that provides half-hourly NO2 concentrations (500 m grid resolution) over smaller spatial extent but accounts for the spatiotemporally varying emissions and wind field. Results. (1) Due to daily varying NO2 concentrations, differences of 6-20 ppb were evident between the residence place and the workplace/school microenvironments. For example, in the Tel-Aviv metropolitan area about 50% of the adults and 66% of the children were exposed to higher concentrations at home than at work/school. (2) Using annual average NOx exposure estimates (to isolate the spatial commute contribution from the temporal daily variation contribution), about 58% of the adults and 70% of children experienced only small (&lt;5 ppb) differences in exposure between the residence and work/school. (3) Commuters (defined as adults (children) that live more than 10 (5) km away from their residence place) and non-commuters had a similar distribution of differences in exposure to NOx between the two microenvironments. Conclusions. Reliable exposure estimation requires accurate information on (a) time-activity patterns (here in terms of spatiotemporal location), and (b) the pollutant spatiotemporal variation.

Visual Voice Activity Detection in the Wild

The visual voice activity detection (V-VAD) problem in unconstrained environments is investigated in this paper. A novel method for V-VAD in the wild, exploiting local shape and motion information appearing at spatiotemporal locations of interest for facial video segment description and the bag of words model for facial video segment representation, is proposed. Facial video segment classification is subsequently performed using the state-of-the-art classification algorithms. Experimental results on one publicly available V-VAD dataset denote the effectiveness of the proposed method, since it achieves better generalization performance in unseen users, when compared to the recently proposed state-of-the-art methods. Additional results on a new unconstrained dataset provide evidence that the proposed method can be effective even in such cases in which any other existing method fails.

Portunes+: Privacy-Preserving Fast Authentication for Dynamic Electric Vehicle Charging

Dynamic contactless charging is an emerging technology for charging electric vehicles (EVs) on the move. For efficient charging and for proper billing, dynamic charging requires secure communication between the charging infrastructure and the EVs that supports very frequent real-time message exchange for EV authentication. In this paper, we propose Portunes+, an authentication protocol for charging pads to authenticate an EVs identity. Portunes+ uses pseudonyms to provide location privacy, allows EVs to roam between different charging sections and receive a single bill, and achieves fast authentication by relying on symmetric keys and on the spatiotemporal location of the EV. We have implemented Portunes+ on RaspberryPi 2 Model B with 900 MHz CPU and 1 GB RAM. Portunes+ allows the EV to generate authentication information within 0.15 ms and allows charging pads to verify the information within 0.11 ms. In comparison, Elliptic Curve Digital Signature Algorithm signature generation and verification take over 9 ms and over 14 ms, respectively.

Where are Universals?

Abstract It is often claimed that realists about universals must be either “platonists”, holding that universals lack spatio-temporal location, or “aristotelians”, asserting that universals are located where their instances are. What’s more, both camps agree that locatedness or unlocatedness is part of the essential nature of universals; consequently, aristotelians say that universals cannot exist un located, and platonists allege that universals cannot be located. Here I argue that the dispute may be resolved by synthesizing the most attractive features of each position. I claim that universals are located at their instances when instantiated by particulars which themselves possess a spatio-temporal location; however, uninstantiated universals exist without location. This has the consequence that it will often be a contingent matter whether a given universal has a location, if the universal is only contingently instantiated. Nevertheless, I claim, objections to this position may be overcome.

Finite-Time Lyapunov Exponents and Lagrangian Coherent Structures in Uncertain Unsteady Flows.

The objective of this paper is to understand transport behavior in uncertain time-varying flow fields by redefining the finite-time Lyapunov exponent (FTLE) and Lagrangian coherent structure (LCS) as stochastic counterparts of their traditional deterministic definitions. Three new concepts are introduced: the distribution of the FTLE (D-FTLE), the FTLE of distributions (FTLE-D), and uncertain LCS (U-LCS). The D-FTLE is the probability density function of FTLE values for every spatiotemporal location, which can be visualized with different statistical measurements. The FTLE-D extends the deterministic FTLE by measuring the divergence of particle distributions. It gives a statistical overview of how transport behaviors vary in neighborhood locations. The U-LCS, the probabilities of finding LCSs over the domain, can be extracted with stochastic ridge finding and density estimation algorithms. We show that our approach produces better results than existing variance-based methods do. Our experiments also show that the combination of D-FTLE, FTLE-D, and U-LCS can help users understand transport behaviors and find separatrices in ensemble simulations of atmospheric processes.

Spatial statistics: Marks, maps, and shapes

ABSTRACTSpatial statistics is concerned with phenomena unfolding in space and possibly also evolving in time, expressing a system of interactions whereby an observation made at a (spatiotemporal) location is informative about observations made at other locations. In general, the interactions are best described probabilistically, rather than deterministically. Spatial scales range from the microscopic (for example, when describing interactions between molecules of a liquid) to planetary (for example, when studying the Earth’s ozone layer) or even larger; temporal scales are similarly varied.Marks indicate objects whose spatial locations are influenced by the presence and nature of other objects nearby: trees of the same or different species in a grove, molecules in a liquid, or galaxies throughout the universe. The statistical models are (marked) spatial point processes.Maps describe the variability of the values of a property across a geographical region. The Ising model of ferromagnetism describes collective properties of atoms arranged in a regular lattice. When mapping the prevalence or the incidence of a disease at the level of counties or parishes, the observations are associated with subsets of a region whose spatial relations are meaningful. Many maps are drawn based on observations made at a finite set of locations distributed either regularly or irregularly throughout a 2D or 3D spatial domain. For example, the mass fraction of uranium in soils and surface sediments across Colorado. Gaussian random functions are a model of choice for such quantities, possibly after re-expression.Shapes arise owing to modulated interactions between surface elements anchored to points in space — “generators” in the nomenclature of Ulf Grenander’s pattern theory. Probability distributions on spaces of generators and on spaces of interactions between them can then be used to describe variations on patterns and to fit shape models.

Cell Tracking Accuracy Measurement Based on Comparison of Acyclic Oriented Graphs

Tracking motile cells in time-lapse series is challenging and is required in many biomedical applications. Cell tracks can be mathematically represented as acyclic oriented graphs. Their vertices describe the spatio-temporal locations of individual cells, whereas the edges represent temporal relationships between them. Such a representation maintains the knowledge of all important cellular events within a captured field of view, such as migration, division, death, and transit through the field of view. The increasing number of cell tracking algorithms calls for comparison of their performance. However, the lack of a standardized cell tracking accuracy measure makes the comparison impracticable. This paper defines and evaluates an accuracy measure for objective and systematic benchmarking of cell tracking algorithms. The measure assumes the existence of a ground-truth reference, and assesses how difficult it is to transform a computed graph into the reference one. The difficulty is measured as a weighted sum of the lowest number of graph operations, such as split, delete, and add a vertex and delete, add, and alter the semantics of an edge, needed to make the graphs identical. The measure behavior is extensively analyzed based on the tracking results provided by the participants of the first Cell Tracking Challenge hosted by the 2013 IEEE International Symposium on Biomedical Imaging. We demonstrate the robustness and stability of the measure against small changes in the choice of weights for diverse cell tracking algorithms and fluorescence microscopy datasets. As the measure penalizes all possible errors in the tracking results and is easy to compute, it may especially help developers and analysts to tune their algorithms according to their needs.