CoMoRea'18 - Program

Monday, March 19, 08:30 - 08:45

S0: Welcome Address

Pascal Hirmer, Heiner Stuckenschmidt, Matthias Wieland
Chairs: Pascal Hirmer (University of Stuttgart, Germany), Matthias Wieland (Universität Stuttgart, Germany)

Monday, March 19, 08:45 - 10:00

S1: Activity Recognition

Chair: Pascal Hirmer (University of Stuttgart, Germany)
08:45 Human Activity Recognition based on Real Life Scenarios
Fadi Al Machot (Alpen Adria University Klagenfurt, Austria); Suneth Ranasinghe (Alpen-Adria-Universität Klagenfurt, Austria); Johanna Plattner (Carinthia University of Applied Sciences, Austria); Nour Jnoub(Universität Wien, Austria)
In Active and Assisted Living (AAL) systems, a major task is to support old people who suffer from diseases such as Dementia or Alzheimer. To provide required support, it is essential to know their Activities of Daily Living (ADL) and support them accordingly. Thus, the accurate recognition of human activities is the foremost task of such an AAL system, especially when non-video/audio sensors are used. It is common that one or more sensors could share or represent a unique activity and the estimation of the most optimal window size for such activity is challenging. Motivated by the powerful learning ability of neural models architectures, this paper proposes to bridge dynamic windowing and Recurrent Neural Networks (RNN), which results in producing the estimated window of sensor events and recognizing the related activity, consequently. The proposed RNN model is trained based on a dynamical systems perspective on weight initialization process. In order to check the overall performance, this approach was tested using the popular CASAS dataset and the newly collected HBMS dataset. Compared to other approaches, the results show a high performance, based on different evaluation metrics. We believe that the proposed windowing approach and RNN model can assist to detect subject independent human activities in smart environment.
09:10 Toward Practical Activity Recognition: Recognizing Complex Activities with Wide Variations
Rabih Younes (Virginia Polytechnic Institute and State University, USA); Mark Jones and Thomas L. Martin (Virginia Tech, USA)
Most activity classifiers focus on recognizing application-specific activities that are mostly performed in a scripted manner, where there is very little room for variation within the activity. These classifiers are mainly good at recognizing short scripted activities that are performed in one way. In reality, especially when considering daily activities, humans perform complex activities in a variety of ways. In this work, we aim to make activity recognition more practical by proposing a novel approach to recognize complex unscripted activities that could be performed in a wide variety of ways. We collect data from 15 subjects performing 8 complex activities and test our approach while analyzing it from different aspects. The results show the validity of our approach. They also show how it performs better than the state-of-the-art approaches that tried to recognize the same activities in a more controlled environment.
09:35 Towards Systematic Benchmarking of Activity Recognition Algorithms
Timo SztylerChristian Meilicke and Heiner Stuckenschmidt (University of Mannheim, Germany)
In this paper we propose a benchmarking framework for evaluating activity recognition methods. We use an ontology for describing activity recognition datasets on the meta-level and propose a fine-grained annotation scheme for activity recognition datasets. Given a method that implements a defined interface, an evaluation client can be used to automatically run experiments on annotated datasets. Our framework helps to find relevant datasets and makes results reproducible by fixing concrete experimental settings. We show how to use the framework and report about a preliminary evaluation experiment.

Monday, March 19, 10:30 - 12:00

S2: Context Prediction

Chair: Matthias Wieland (Universität Stuttgart, Germany)
10:30 Correlation-Based Pre-Filtering for Context-Aware Recommendation
Zahra Vahidi Ferdousi and Dario Colazzo (University of Paris-Dauphine, France); Elsa Negre (Université Paris Dauphine, LAMSADE, France)
With the increasing use of connected devices and IoT, users' contextual information is more and more available and used in different information systems. One of the domains where the use of contextual information is promising is that of recommendation. As a matter of fact, context-aware recommender systems (CARSs) have demonstrated that taking contextual information about users into account can improve the effectiveness of recommendation, by generating more relevant recommendations to the users in their specific contextual situation. In this paper we propose a new context representation and approach to integrate this kind of information into a recommender system. We make a strong representation of the context, based on the influence of context on ratings, calculated using the Pearson Correlation Coefficient. We do a prefiltering recommendation based on this representation. Our evaluations demonstrate that our approach can outperforms the state of the art.
11:00 Predicting Contextual Influences on App Usage from a Rational Model of Time Allocation
Robert Edge and Dominic Mussack (University of Minnesota, USA); Matthias Böhmer (TH Köln, Germany); Paul Schrater (University of Minnesota, USA)
Mobile devices have proven to be a transformative tool that help users perform a variety of everyday tasks. However, they also have tremendous potential to disrupt productive and desired time allocation, facilitating time-squandering through self interruptions of workflow and undesired task switching through distracting apps. Existing research has identified a variety of context variables which help predict the next app selected, but seldom give treatment to the pattern of app usage durations essential to understanding time allocation. Here we take a psychological computing approach to identify the key environmental factors that increase risk of early termination through unwanted switching. Using a task foraging model for time allocation, we construct an integrated measure of the background factors increasing switching temptation, and show that these can be converted into a computable measure of decision context that strongly impacts app duration. The foraging model gives new insight into the structural factors that promote task persistence and predict switch temptations, and suggests new ways to design productive environments.
11:30 Purpose-of-Visit-Driven Semantic Similarity Analysis on Semantic Trajectories for Enhancing The Future Location Prediction
Antonios Karatzoglou (Robert Bosch GmbH & Karlsruhe Institute of Technology (KIT), Germany); Dominik Koehler (Karlsruhe Institute of Technology, Germany); Michael Beigl (KIT & TECO, Germany)
The number of people that are using or are even dependent on Location Based Services (LBS) is growing rapidly every year. Recently, in response to the trend towards offering timely and tailored to the users services, providers rely increasingly on forward-looking algorithms. Context and especially location prediction plays therefore a key role in the respect of LBS. At the same time, there exists a relative new but promising research work on semantic-enhanced location prediction, able to overcome some drawbacks that characterize conventional non-semantic systems. However, the majority constraints itself to fixed static semantical constructs without taking the user's current situation into account. In this work, we present a dynamic framework, that considers explicitly the purpose of visit in order for the prediction performance to be elevated. Our framework is hybrid and combines both a data-driven, as well as a knowledge-driven model. Moreover, we de ne a Purpose-of-Visit Driven Semantic Similarity (PoVDSS) and use it as an intelligent fusing component. We conducted a user study to evaluate our approach on real data and compared it with two well known semantic and non-semantic frameworks. Our evaluation shows that our approach yields to results up to 60% f-score.

Monday, March 19, 13:15 - 15:00

S3: Internet of Things

Chair: Pascal Hirmer (University of Stuttgart, Germany)
13:15 An Approach for CEP Query Shipping to Support Distributed IoT Environments
In recent years, the amount of data highly increases. Deriving information and, consequently, knowledge from this data leads to huge benefits. To realize this, oftentimes Complex Event Processing is employed. Usually, current solutions process data on monolithic IT infrastructures. However, for the emerging Internet of Things paradigm, this is not adequate, because high efficiency is of vital importance. To achieve this, distributed data processing with short communication paths and reduced network traffic need to be enabled. In this paper, we introduce an approach for CEP query shipping to support distributed Internet of Things environments.
13:50 Inferring Availability for Communication in Smart Homes Using Context
Julien Cumin (Université Grenoble Alpes & Orange Labs, France); Grégoire Lefebvre and Fano Ramparany (Orange Labs, France); James Crowley (Université Grenoble Alpes, France)
This paper presents a technique for inferring the availability of people to receive communications based on their current situation. This technique uses a context model that associates situations with learned preferences for communications. Situations are represented as a tuple composed of identity, time, place, activity, correspondent, and communications modality. A place-based activity recognition technique is used to recognize the current activity from sensor information. Availability for communications is learned from history of the occupant's preferences of availability for each situation. The system is demonstrated using a dataset of availability preferences recorded from the occupant of an instrumented apartment over a period of 4 weeks. Performance of the system is compared under various assumptions of independence of availability from some of the context elements. The paper is completed with a discussion of how such techniques can be used to construct an intelligent communications assistant for smart home services.
14:25 Modeling and Reasoning with ProbLog: An Application in Recognizing Complex Activities
Timo Sztyler (University of Mannheim, Germany); Gabriele Civitarese (University of Milan, Italy); Heiner Stuckenschmidt (University of Mannheim, Germany)
Smart-home activity recognition is an enabling tool for a wide range of ambient assisted living applications. The recognition of ADLs usually relies on supervised learning or knowledge-based reasoning techniques. In order to overcome the well-known limitations of those two approaches and, at the same time, to combine their strengths to improve the recognition rate, many researchers investigated Markov Logic Networks (MLNs). However, MLNs require a non-trivial effort by experts to properly model probabilities in terms of weights. In this paper, we propose a novel method based on ProbLog. ProbLog is a probabilistic extension of Prolog, which allows to explicitly define probabilistic facts and rules. With respect to MLN, the inference mode of ProbLog is based on the closed-world assumption and it has faster response times. We propose a simple and flexible ProbLog model, which we exploit to recognize complex ADLs in an online fashion. Considering a dataset with 21 subjects, our results show that our method reaches high F-measure (83%). Moreover, we also show that the response time of ProbLog is satisfying for real-time applications.

Monday, March 19, 15:30 - 17:00

S4: Privacy, Safety and Reasoning

Chair: Matthias Wieland (University of Stuttgart, Germany)
15:30 How a Pattern-based Privacy System Contributes to Improve Context Recognition
As Smart Devices have access to a lot of user-preferential data, they come in handy in any situation. Although such data - as well as the knowledge which can be derived from it - is highly beneficial as apps are able to adapt their services appropriate to the respective context, it also poses a privacy threat. Thus, a lot of research work is done regarding privacy. Yet, all approaches obfuscate certain attributes which has a negative impact on context recognition and thus service quality. Therefore, we introduce a novel access control mechanism called PATRON. The basic idea is to control access to information patterns. For instance, a person suffering from diabetes might not want to reveal his or her unhealthy eating habit, which can be derived from the pattern "rising blood sugar level" -> "adding bread units". Such a pattern which must not be discoverable by some parties (e.g., insurance companies) is called private pattern whereas a pattern which improves an app's service quality is labeled as public pattern. PATRON employs different techniques to conceal private patterns and, in case of available alternatives, selects the one with the least negative impact on service quality, such that the recognition of public patterns is supported as good as possible.
16:00 The Wireless Seat Belt Requirements, Experiments, and Solutions for Pedestrian Safety
Marek BachmannMichel Morold and Klaus David (University of Kassel, Germany); Patrick Henkel (Technische Universität München, Germany)
The World Health Organization in its latest report on road safety states that 22% of all road traffic deaths comprise pedestrians, approximately 275,000 pedestrians worldwide. Based on an overview of typical accident scenarios, we argue how an optimized pedestrian safety system should look like. We provide a comprehensive analysis of the accuracy requirements of position, direction, and speed for such an optimized system. Next, we experimentally show what can be done with current smartphones in terms of these requirements. We introduce and evaluate two approaches using context information to enable "off-the-shelf" smartphones to reach the required accuracy requirements. This gives us the confidence, that our innovative pedestrian safety approach, which we call Wireless Seat Belt, can save lives.
16:30 Custom Dual Transportation Mode Detection by Smartphone Devices Exploiting Sensor Diversity
Claudia Carpineti (University of Bologna, Italy); Vincenzo Lomonaco (DISI, Italy); Luca BedogniMarco Di Felice and Luciano Bononi (University of Bologna, Italy)
Making applications aware of the mobility experienced by the user can open the door to a wide range of novel services in different use-cases, from smart parking to vehicular traffic monitoring. In the literature, there are many different studies demonstrating the theoretical possibility of performing Transportation Mode Detection (TMD) by mining smartphones embedded sensors data. However, few of them provide details on the benchmarking process and on how to implement the detection process in practice. In this study, we provide guidelines and fundamental results that can be useful for both researcher and practitioners aiming at implementing a working TMD system. The guidelines consist of three main contributions. First, we detail the construction of a training dataset, gathered by heterogeneous users and including five different transportation modes; the dataset is made available to the research community as reference benchmark. Second, we provide an in-depth analysis of the sensor-relevance for the case of Dual TDM, which is required by most of mobility-aware applications. Third, we investigate the possibility to perform TMD of unknown users/instances not present in the training set and we compare with state-of-the-art Android APIs for activity recognition.

Monday, March 19, 17:00 - 17:30

S5: Discussion Plenum

Pascal Hirmer, Matthias Wieland
Chairs: Pascal Hirmer (University of Stuttgart, Germany), Matthias Wieland (Universität Stuttgart, Germany)