Event Recognition for Behavior Measurement, Intelligent Resource Management, and Boyond

WORKSHOP

Date: Friday, August 27
Time: 10:00-12:30
Location: Ernst
Chair: Jobst Löffler (Fraunhofer Institute for Intelligent Analysis and Information Systems IAIS, Germany). Ben Loke (Noldus Information Technology bv,The Netherlands). Jens Pottebaum (University of Paderborn, Germany)

Complex event processing, real-time data analysis, and event-driven software systems are highly relevant topics for various research communities as well as for enterprises dealing with enormous amounts of business data. New approaches to extracting information, transforming it into knowledge, and acting on it have to be found. With this workshop we would like to offer an opportunity for researchers, practitioners and technology providers to discuss techniques for detecting, tracking, and processing events that can be applied in behavior measurement. Further on the utilization of behavior measurement results, e.g. intelligent resource management and corresponding feedback to measurement processes will be discussed. During the workshop the underlying concepts of event-driven information management will be introduced.

As a practical example and starting point for discussion, methods, tools, and applications which have been developed within the EC funded FP7 project PRONTO (www.ict-pronto.org) will be demonstrated. PRONTO emphasizes the role of event recognition in intelligent resource management. The project proposes a methodology for analyzing data from various sources to extract useful information in the form of events and fusing it to actionable information. The PRONTO approach will be evaluated in two case studies in the domains of emergency rescue operations and city transport  management.

Topics
Topics of interest include, but are not limited to:

• Algorithms for real-time event recognition
• Multi-modal/fusion techniques for sensor data and event streams
• Benchmarks, performance evaluations, and test beds for the application of event recognition to behavior measurement
• Indicators for the detection of events in complex real world systems
• User interaction and interfaces for event navigation, browsing, and management
• Utilization of behavior measurement in decision making processes and lessons learned
• Domain-specific deployments of event-based systems

During the workshop presentations will be given that summarize current research in the field of Event Recognition for Behavior Measurement, Intelligent Resource Management, and beyond. They open the discussion on bringing together different approaches from different domains to strengthen research efforts. In detail the contributions will focus on the following activities:

Alex Lechleuthner and Ompe-Aimé Mudimu (The Institute of Rescue Engineering and Emergency Management of Cologne University of Applied Sciences, Germany) are involved with the laboratory (Benedikt Weber, Frederik Schütte and Benjamin Käser ) for major incidents in the area of training observation and infield-research. The Standard-Exercise-System (SES) allows scientific research and evaluation with infield tests. With this system, data is aggregated using processes that enable a fundamental examination of safety and security systems. The first step is a parameter definition to identify the measurement points. Furthermore, specific parameters and process components are changed or replaced in order to measure improvements and to formulate user specifications. The
measurements are based on a Tele Voting system. This system was adapted and expanded for use in the area of safety and security. Essentially, data is collected in three different modes: live mode, presentation mode and offline mode. In the future, this system is supplemented by a local positioning system.

Gertjan Burghouts, Maaike Lousberg and Judith Dijk (TNO Defense, Security & Safety, Den Haag, The Netherlands) focuses on human behaviour that precedes unwanted situations. Examples of these unwanted situations are everyday crime, escalating events and demonstrations, aggression in the cities, public transport, acts of violence, the destruction of windows, cables, and the entering of critical infrastructures and private properties. To ensure public safety and security, an adequate awareness of the situation plays a critical role for the professional. Video, audio and sensor based surveillance is a critical asset to achieve such awareness. TNO analyzes the situation and the behaviour of the people in the scene. First, deviant behaviors are identified using knowledge of psychology. Then, sensor and contextual data is collected from the available sources and analyzed in two sequential steps. In the first step, the software detects behavioral characteristics that may hint at unwanted situations. In the second step, the software interprets the detected behaviors, to assess whether unwanted situations are probable.

Jeroen van Rest (TNO, Security & Safety, Den Haag, The Netherlands) and his team argue that designing systems for human behaviour observation is difficult. Subjects might not cooperate, or might even be hostile towards the system. The observation system itself might influence the observation. The personal data that is gathered falls under privacy laws and regulations which complicate the design. Yet human behaviour observation systems (HBOS) are being applied in many domains, such as security, mobility, care, entertainment and science. Due to technological advancements, mainly driven by Moore’s Law and the human drive to codify knowledge into systems, the technological building blocks of HBOS show exponential increasing capabilities. In all developed countries the legislative branch is responding by updating privacy laws, but social new media tempts individuals to register personal data out of free will. In this complex and dynamic setting, the designers of HBOS have their challenges cut out for them. The team argues that design patterns exist for HBOS. Unlike best practices, these patterns are independent of culture, application domain or technological development. A design pattern is primarily identified by a unique name, and described by a challenge and a solution for that challenge. It can be explained by showing the recurring pattern in different example implementations. Design patterns are well known in software engineering, but they can be generalized to other aspects, such as the physical or temporal dimensions of a system.

The workshop can be of interest for computer scientists, mathematicians, AI experts, as well as behavioral biologists and psychologists. We would like to initiate a lively discussion on the topics listed above and no specific background knowledge is required for participation. Attending this workshop will provide the participants with a more refined view on the advantages and disadvantages, with respect to applicability as well as output, of event-driven approaches.