Unusual Event Detection Research Articles

Using marine mammal necropsy data in animal health surveillance: the case of the harbor porpoise in the Southern North Sea

Rapid changes of marine ecosystems resulting from human activities and climate change, and the subsequent reported rise of infectious diseases in marine mammals, highlight the urgency for timely detection of unusual health events negatively affecting populations. Studies reporting pathological findings in the commonly stranded harbor porpoise (Phocoena phocoena) on North Atlantic coastlines are essential to describe new and emerging causes of mortality. However, such studies often cannot be used as long-term health surveillance tools due to analytical limitations. We tested 31 variables gained from stranding-, necropsy-, dietary- and marine debris data from 405 harbor porpoises using applied supervised and unsupervised machine learning techniques to explore and analyze this large dataset. We classified and cross-correlated the variables and characterized the importance of the different variables for accurately predicting cause-of-death categories, to allow trend assessment for good conservation decision. The variable ‘age class’ seemed most influential in determining cause-of-death categories, and it became apparent that juveniles died more often due to acute causes, including bycatch, grey-seal-predation and other trauma, while adults of infectious diseases. Neonates were found in summer, and mostly without prey in their stomach and more often stranded alive. The variables assigned as part of the external examination of carcasses, such as imprints from nets and lesions induced by predators, as well as nutritional condition were most important for predicting cause-of-death categories, with a model prediction accuracy of 75%. Future porpoise monitoring, and in particular the assessment of temporal trends, should predominantly focus on influential variables as determined in this study. Pathogen- and contaminant assessment data was not available for all cases, but would be an important step to further complete the dataset. This could be vital for drawing population-inferences and thus for long-term harbor porpoise population health monitoring as an early warning tool for population change.

Early detection of wildlife morbidity and mortality through an event-based surveillance system.

The ability to rapidly detect and respond to wildlife morbidity and mortality events is critical for reducing threats to wildlife populations. Surveillance systems that use pre-diagnostic clinical data can contribute to the early detection of wildlife morbidities caused by a multitude of threats, including disease and anthropogenic disturbances. Here, we demonstrate proof of concept for use of a wildlife disease surveillance system, the ‘Wildlife Morbidity and Mortality Event Alert System’, that integrates pre-diagnostic clinical data in near real-time from a network of wildlife rehabilitation organizations, for early and enhanced detection of unusual wildlife morbidity and mortality events. The system classifies clinical pre-diagnostic data into relevant clinical classifications based on a natural language processing algorithm, generating alerts when more than the expected number of cases is recorded across the rehabilitation network. We demonstrated the effectiveness and efficiency of the system in alerting to events associated with both common and emerging diseases. Tapping into this readily available unconventional general surveillance data stream offers added value to existing wildlife disease surveillance programmes through a relatively efficient, low-cost strategy for the early detection of threats.

IA-SSLM: Irregularity-Aware Semi-Supervised Deep Learning Model for Analyzing Unusual Events in Crowds

Analyzing unusual events is significantly important for video surveillance to ensure people safety. These events are characterized by irregular patterns that do not conform to the expected behavior in the surveillance scenes. We present a novel irregularity-aware semi-supervised deep learning model (IA-SSLM) for detection of unusual events. While most existing works depend on the availability of large amount of labeled data for training, our proposed method utilizes a semi-supervised deep model to automatically learn feature representations from limited number of labeled data samples. Our method extracts meaningful information from both labeled and unlabeled data during the training stage to improve the performance. For this purpose, we explore the concept of consistency regularization and entropy minimization to output confident predictions on unlabeled data. For experimental analysis, we consider various standard and diverse datasets. The results show that our IA-SSLM method outperforms several reference methods using different performance metrics.

Real-time unusual user event detection algorithm fusing vision, audio, activity, and dust patterns

According to the statistics examined by the National Safety Council, Injury Facts in 2017, a significant number of preventable injury-related deaths occurred in home and indoor public areas, and the rate of preventable deaths occurring indoors has increased by 156% since 1999. Indoor smart Internet of Things devices such as security cameras, intelligent speakers, smartphones, and air cleaners are being utilized to seek help during dangerous or emergency situations in indoor settings. Theses Internet of Things devices can also assist single-person household rescues during emergency situations where no one is present for assistance. We propose a real-time algorithm to detect unusual user events which may help in reducing the rate of people dying alone and remaining undiscovered for a long period of time. We designed and developed unusual user behavior patterns related to vision, audio, dust, and activity via Internet of Things sensors and fused these patterns to improve the performance accuracy. We evaluated the proposed individual pattern algorithms and the fusion method through the data collected in indoor smart environments.

Intelligent Pattern Recognition Algorithms based on Dust, Vision and Activity Sensors for User Unusual Event Detection

Intelligent Pattern Recognition Algorithms based on Dust, Vision and Activity Sensors for User Unusual Event Detection

Assessing safety of Ontario's publicly funded MMR and MMRV immunization programs, 2012 to 2016.

The combined measles, mumps, rubella (MMR) and measles, mumps, rubella, and varicella (MMRV) vaccines are part of Ontario's routine immunization schedule. To assess adverse events following immunization (AEFIs) reported in Ontario following administration of MMR and MMRV vaccines between 2012 and 2016. Reports of AEFIs were extracted from the provincial surveillance database on May 9, 2017. Events were grouped by provincial surveillance definitions. Reporting rates were calculated using provincial population estimates or net doses distributed as the denominator. A serious AEFI is defined as an AEFI that resulted in an in-patient hospitalization or death. Overall, 289 AEFIs were reported following administration of MMR (n=246) or MMRV (n=43) vaccines, for annualized reporting rates of 16.6 and 8.8 reports per 100,000 distributed doses, respectively. The highest age-specific reporting rate was in children aged 1 to 3 years for MMR (7.7 per 100,000 population) and children aged 4 to 9 years for MMRV (0.8 per 100,000 population). Systemic reactions were the most frequently reported event category, while rash was the most frequently reported event for both vaccines. There were 22 serious AEFIs, 19 following MMR and 3 following MMRV (1.3 and 0.6 per 100,000 doses distributed, respectively). Our assessment found a low reporting rate of adverse events following MMR and MMRV vaccines in Ontario. No safety concerns were identified. Our findings are consistent with the safety profiles of these vaccines. Continued monitoring of vaccine safety is necessary to maintain timely detection of unusual postvaccine events and public confidence in vaccine safety.

Integrated Disease Surveillance and Response (IDSR) in Malawi: Implementation gaps and challenges for timely alert.

ObjectiveThe recent 2014 Ebola Virus Disease (EVD) outbreaks rang the bell to call upon global efforts to assist resource-constrained countries to strengthen public health surveillance system for early response. Malawi adopted the Integrated Disease Surveillance and Response (IDSR) strategy to develop its national surveillance system since 2002 and revised its guideline to fulfill the International Health Regulation (IHR) requirements in 2014. This study aimed to understand the state of IDSR implementation and differences between guideline and practice for future disease surveillance system strengthening.MethodsThis was a mixed-method research study. Quantitative data were to analyze completeness and timeliness of surveillance system performance from national District Health Information System 2 (DHIS2) during October 2014 to September 2016. Qualitative data were collected through interviews with 29 frontline health service providers from the selected district and 7 key informants of the IDSR system implementation and administration at district and national levels.FindingsThe current IDSR system showed relatively good completeness (73.1%) but poor timeliness (40.2%) of total expected monthly reports nationwide and zero weekly reports during the study period. Major implementation gaps were lack of weekly report and trainings. The challenges of IDSR implementation revealed through qualitative data included case identification, compiling reports for timely submission and inadequate resources.ConclusionsThe differences between IDSR technical guideline and actual practice were huge. The developing information technology infrastructure in Malawi and emerging mobile health (mHealth) technology can be opportunities for the country to overcome these challenges and improve surveillance system to have better timeliness for the outbreaks and unusual events detection.

A Low-Power Vision System With Adaptive Background Subtraction and Image Segmentation for Unusual Event Detection

This paper presents a smart ultra-low power vision system targeted to video surveillance applications. The sensor embeds a low-level image processing technique that autonomously detects unusual events occurring in the scene, relying on adaptive background subtraction. The resulting binary image is then directly segmented by an FPGA, which triggers the higher layer of processing, transferring only aggregate feature information. The on-board processing relieves the rest of the vision system from expensive computation. The $104\times 104$ pixels vision chip consumes $80~\mu \text{W}$ at 30 frames/s, while segmentation dramatically cuts down the amount of data to be transferred, resulting in an extremely low-power system suitable for embedded applications.

Personalized Unusual Event Detection Algorithm at Smart Home via Daily Activity and Vision Pattern

Personalized Unusual Event Detection Algorithm at Smart Home via Daily Activity and Vision Pattern

Variational Bayesian Learning of Generalized Dirichlet-Based Hidden Markov Models Applied to Unusual Events Detection.

Learning a hidden Markov model (HMM) is typically based on the computation of a likelihood which is intractable due to a summation over all possible combinations of states and mixture components. This estimation is often tackled by a maximization strategy, which is known as the Baum-Welch algorithm. However, some drawbacks of this approach have led to the consideration of Bayesian methods that add a prior over the parameters in order to work with the posterior probability and the marginal likelihood. These approaches can lead to good models but to the cost of extremely long computations (e.g., Markov Chain Monte Carlo). More recently, variational Bayesian frameworks have been proposed as a Bayesian alternative that keeps the computation tractable and the approximation tight. It relies on the introduction of a prior over the parameters to be learned and on an approximation of the true posterior distribution. After proving good standing in the case of finite mixture models and discrete and Gaussian HMMs, we propose here to derive the equations of the variational learning of the Dirichlet mixture-based HMM, and to extend it to the generalized Dirichlet. The latter case presents several properties that make the estimation more accurate. We prove the validity of this approach within the context of unusual event detection in public areas using the University of California San Diego data sets. HMMs are trained over normal video sequences using the typical Baum-Welch approach versus the variational one. The variational learning leads to more accurate models for the detection and localization of anomaly, and the general HMM approach is shown to be versatile enough to handle the detection of various synthetically generated tampering events.

Deep-Structured Event Modeling for User-Generated Photos

Vision-based event analysis is difficult because of the following challenges. The first challenge is intraclass variation. Photos uploaded by users are sparsely sampled visual appearances of an event over time. Thus, each photo may only capture a single object or scene of a specific complex event. The second challenge is interclass confusion. Photos related to different events may contain similar objects or scenes. Third, unusual events are characterized by scarcity, and only a few samples are available for use in learning event patterns. In this paper, by considering the photo timestamp, we propose a structured event modeling (SEM) framework for event analysis that exploits the temporal information of visual features and event classes in a photo sequence. Specifically, the temporal event patterns of the photo sequence and the relationships of different photos are jointly learned using deep neural networks (convolutional neural networks and recurrent neural networks) and a conditional random field. We evaluate the proposed SEM framework in two applications: multiclass event recognition and unusual event detection in photo sequences. The results of extensive experiments performed on a public event recognition dataset and a collected unusual event dataset demonstrate the effectiveness of the proposed method.

An Interactive Visual Analytics Platform for Smart Intelligent Transportation Systems Management

The reduction of road congestion requires intuitive urban congestion-control platforms that can facilitate transport stakeholders in decision making. Interactive ITS visual analytics tools can be of significant assistance, through their real-time interactive visualizations, supported by advanced data analysis algorithms. In this paper, an interactive visual analytics platform is introduced that allows the exploration of historical data and the prediction of future traffic through a unified interactive interface. The platform is backed by several data analysis techniques, such as road behavioral visualization and clustering, anomaly detection, and traffic prediction, allowing the exploration of behavioral similarities between roads, the visual detection of unusual events, the testing of hypotheses, and the prediction of traffic flow after hypothetical incidents imposed by the human operator. The accuracy of the prediction algorithms is verified through benchmark comparisons, while the applicability of the proposed toolkit in facilitating decision making is demonstrated in a variety of use case scenarios, using real traffic and incident data sets.

Sensing the city with Instagram: Clustering geolocated data for outlier detection

Early detection of unusual events in urban areas is a priority for city management departments, which usually deploy specific complex video-based infrastructures typically monitored by human staff. However, and with the emergence and quick popularity of Location-based social networks (LBSNs), detecting abnormally high or low number of citizens in a specific area at a specific time could be done by an expert system that automatically analyzes the public geo-tagged posts. Our approach focuses exclusively on the location information linked to these posts. By applying a density-based clustering algorithm, we obtain the pulse of the city (24 h–7 days) in a first training phase, which enables the detection of outliers (unexpected behaviors) on-the-fly in an ulterior test or monitoring phase. This solution entails that no specific infrastructure is needed since the citizens are the ones who buy, maintain, carry the mobile devices and freely disclose their location by proactively sharing posts. Besides, location analysis is lighter than video analysis and can be automatically done. Our approach was validated using a dataset of geo-tagged posts obtained from Instagram in New York City for almost six months with good results. Actually, not only all the already previously known events where detected, but also other unknown events where discovered during the experiment.

Bayesian prospective detection of small area health anomalies using Kullback-Leibler divergence.

Early detection of unusual health events depends on the ability to rapidly detect any substantial changes in disease, thus facilitating timely public health interventions. To assist public health practitioners to make decisions, statistical methods are adopted to assess unusual events in real time. We introduce a surveillance Kullback-Leibler measure for timely detection of disease outbreaks for small area health data. The detection methods are compared with the surveillance conditional predictive ordinate within the framework of Bayesian hierarchical Poisson modeling and applied to a case study of a group of respiratory system diseases observed weekly in South Carolina counties. Properties of the proposed surveillance techniques including timeliness and detection precision are investigated using a simulation study.

Proportional data modeling with hidden Markov models based on generalized Dirichlet and Beta-Liouville mixtures applied to anomaly detection in public areas

The recent and rapid deployment of CCTV cameras in public areas invokes the need for a capability to assist human operators in the real-time detection of threats and anomalous events. The main difficulty in this regard is the rare and unpredictable nature of the events to be detected. A typical strategy consists in modeling situations considered as normal and detecting outliers to this model. In this paper, we theoretically derive two variants of the hidden Markov model for proportional data modeling and assess their performance for unusual event detection in several public surveillance situations. Proportional data arise in a number of pattern recognition applications and classic hidden Markov models are not well adapted so far for their specific processing. Our models are based on the use of the generalized Dirichlet and Beta-Liouville distributions as emission probability functions and yield enhanced performance compared to the use of the Dirichlet distribution.

Statistical monitoring-based alarming systems in modeling the AIDS epidemic in the United States, 1985-2011.

Better decisions for the control of HIV/AIDS and other infectious diseases require better information. The large amount of available public health data makes it possible to extract such information to monitor and predict significant disease events in disease epidemic. The detection of unusual events often involves a combination of a forecasting and a decision mechanism assessing the extent to which an observed event differs significantly from a forecast event. A number of methods and models have been proposed to monitor the trend of infectious disease and to detect unusual events. Although these existing methods and models are useful, many new issues remain to be addressed, including the complicated data structure and the infectious disease dynamics. To overcome these issues, we introduced the statistical tool using statistical process control, and proposed a new method under that framework. In this paper, we first reviewed the most commonly used methods and models, including the historical limit method, the time series analysis, the hidden Markov models, and the process control charts. Then, we further discussed issues with the current available methods. We proposed a new method using statistical process control. A major feature of the new method is that it prospectively monitors the disease incidence using sequentially collected data over time. It also takes into account a wide variety of longitudinal patterns and possible autocorrelation in the data. We test this novel method with the recorded data of the number of AIDS cases in different states of US from 1985 to 2011. The results show that our new method is effective in detecting and predicting the time trends of AIDS epidemic for individual states and for US as a whole. Although AIDS data are used in our demonstration, this method can be used for monitoring other infectious diseases.

Using daily syndrome-specific absence data for early detection of school outbreaks: a pilot study in rural China

ObjectivesThis paper describes and preliminarily evaluates the usefulness of the daily syndrome-specific absenteeism surveillance system (DSSASS) as an early warning system of school outbreaks in rural China. Study designWe conducted an experimental study in rural areas of Hubei Province from September 19, 2011 to December 31, 2011. MethodsNine public elementary schools from two counties were selected as pilot sentinel schools. Daily monitoring data of the absent date and reason, sex, age and class of each absent student was collected and entered into a web database. Reported data were checked daily and field investigation was carried out when there was abnormal absentee aggregation. Descriptive analysis and preliminary evaluation were then conducted after the pilot study. ResultsThe findings showed that the total average of daily absenteeism rate was 3%, and the absenteeism rate differed by county, school level and grade level. The daily absenteeism rate in illness absentees was highest (2.74%), followed by business absentees (0.13%) and injury absentees (0.09%). The total timeliness report rate was 64.84% and the total incident report rate was 29.22%. One varicella outbreak and one influenza B outbreak were identified, but neither of them was detected by China Information System for Diseases Control and Prevention (CISDCP). The study shows syndrome-specific absenteeism data would be useful for early detection of unusual public health events or outbreaks in school. However, more efforts are needed to enhance the quality of surveillance data, and longer follow-up and more analysis are required to evaluate the system comprehensively. Our study might provide useful experience and evidence for other developing regions or counties establishing similar systems.

Testing of Event Detection Lateness in Wireless Sensor Networks

A wireless sensor network (WSNs) technology determines physical properties of the environment through detection of unusual events. The sensor consists of small mini computers powered by batteries and use radio communication for the exchange of information it gathers. The sensors are able to collect the data and exchange the information with all the nodes that were affected and triggered during the event detection. Although the technology has been in use for a period of time it has been noted that there is lateness of the event detection by the technology. Therefore the essence of this paper is to describe the factors affecting the lateness detection of events by WSNs. From the experiment it is evident that varying distance of the sensor nodes to the base station and the transmission power affect the packet data transmission efficiency and thus lateness detection of events. Power level at 5 dBm (maximum level) ensures 90% of the event detected is communicated effectively to the base station whereas minimum power level at -10 dBm allows ineffective detection of events. At -10dBm the base station node could not detect a transmitted data packet at distances beyond 30 feet. Thus the application of any sensor node with the transmitting power at -10 dBm will only be effective at the distance less than 10ft. Therefore, distance of the sensor nodes and the transmission power of the WSNs affect the data packet delivery performance and hence lateness of event detection.

Personalized behavior pattern recognition and unusual event detection for mobile users

Mobile phones have become widely used for obtaining help in emergencies, such as accidents, crimes, or health emergencies. The smartphone is an essential device that can record emergency situations...

Syndromic Surveillance Based on Emergency Visits: A Reactive Tool for Unusual Events Detection

During the past years, syndromic surveillance based on emergency visits allowed to monitor seasonal epidemics in Reunion Island and to measure their health impact. Nevertheless, its ability to detect minor unusual events still had to be demonstrated. Between 2011 and 2012, several small scale unusual health events were detected (autochthonous cases of measles, family foodborne illness etc.), leading to health alerts and requiring immediate control measures.