Hedda Rahel Schmidtke

ActiServ: Activity Recognition Service for mobile phones

Smart phones have become a powerful platform for wearable context recognition. We present a service-based recognition architecture which creates an evolving classification system using feedback from the user community. The approach utilizes classifiers based on fuzzy inference systems which use live annotation to personalize the classifier instance on the device. Our recognition system is designed for everyday use: it allows flexible placement of the device (no assumed or fixed position), requires only minimal personalization effort from the user (1–3 minutes per activity) and is capable of detecting a high number of activities. The components of the service are shown in an evaluation scenario, in which recognition rates up to 97% can be achieved for ten activity classes.

Using web service gateways and code generation for sustainable IoT system development

Wireless Sensing and Radio Identification systems have undergone many innovations during the past years. This has led to short product lifetimes for both software and hardware compared to classical industries. However, especially industries dealing with long-term support of products, e.g. of industrial machinery, and product lifetime of 40+ years may especially profit from an Internet of Things. Motivated by a practical industrial servicing use case this paper shows how we hope to make equally sustainable IoT solutions by employing a model driven software development approach based on code generation for multi-protocol web service gateways.

The House Is North of the River: Relative Localization of Extended Objects

This article presents a proposal to directional localization of extended objects based on an analysis of the underlying geometric structures. Many existing accounts on the directional localization of objects are based on representing the objects as points. These approaches provide good inferential properties, but are unsuitable to be combined with topological information about the regions occupied by objects. The approach presented here describes directional localization relations on the basis of representing objects as extended regions. Directional localization of an extended region based on a system of sectors and the establishment of a system of sectors based on an extended region are considered independently and the solutions are finally combined. The investigations yield a geometric characterization of directional localization relations in twodimensional space that can be applied to objects represented by regions or curves.

Towards Ontology-Based Formal Verification Methods for Context Aware Systems

Pervasive computing systems work within, and rely on, a model of the environment they operate in. In this respect, pervasive computing systems differ from other distributed and mobile computing systems, and require new verification methods. A range of methods and tools exist for verifying distributed and mobile concurrent systems, and for checking consistency of ontology-based context models. As a tool for verifying current pervasive computing systems both are not optimal, since the former cover mainly tree-based location models, whereas the latter are not able to address the dynamic aspects of computing systems. We propose to formally describe pervasive computing systems as distributed concurrent systems operating on the background of a mereotopological context model.

A geometric agent following route instructions

We present the model of a Geometric Agent that can navigate on routes in a virtual planar environment according to natural-language instructions presented in advance. The Geometric Agent provides a new method to study the interaction between the spatial information given in route instructions and the spatial information gained from perception. Perception and action of the Geometric Agent are simulated. Therefore, the influence of differences in both linguistic and perceptual skills can be subject to further studies employing the Geometric Agent. The goal of this investigation is to build a formal framework that can demonstrate the performance of specific theories of the interpretation of natural-language in the presence of sensing. In this article, we describe the main sub-tasks of instructed navigation and the internal representations the Geometric Agent builds up in order to carry them out.

Granularity as a parameter of context

Spatial and temporal granularity can be understood as parameters of context restricting the set of accessible objects in a context. Starting from the idea that this selection process depends to a large extent on the relation between the grain-size of the context and the local extension of the objects, the granularity of a context is in this article formalised as a class of possible sizes in the context. This formalisation is shown to be in accordance to well-known mathematical foundations on perceptual classification. An example for the case of temporal granularity illustrates how the introduction of new elements into a context may result in a more or less smooth shifting of the granularity leading to a classification of four different types of change of granularity. The results can be applied in a wide range of fields, e.g. in research on contextual reasoning and natural language understanding.

A novel micro-vibration sensor for activity recognition: Potential and limitations

This paper researches the potential of a novel ball switch as a wearable vibration sensor for activity recognition. The ball switch is available as a commercial, off-the-shelf sensor and is unique among such sensors due to its miniaturized design and the low mass of the ball. We present a detailed analysis of the physical properties of the sensor as well as a recommendation for circuit design, sampling method and a feature generation algorithm for activity recognition. The analysis reveals that it is sensitive to vibrations between 1.5 kHz and 8 kHz, where the acceleration sensor is responsive below 1.6 kHz. Furthermore, the ball switch is substantially cheaper (3x), smaller (2x) and uses less power (50x) than an accelerometer based system, but delivers less information. We also present the results of a case study in activity recognition done in parallel with an acceleration sensor using 5 subjects and 8 different activities. It shows that the ball switch can increase recognition rates when added to an accelerometer-based system, demonstrating that it can sample activity-pertinent information which an accelerometer can not. We conclude that this ball switch can be used to recognize high-frequency activity components and effectively improve recognition rates while representing a very low cost sensor in terms of price, device size and power consumption.

An extensible modular recognition concept that makes activity recognition practical

In mobile and ubiquitous computing, there is a strong need for supporting different users with different interests, needs, and demands. Activity recognition systems for context aware computing applications usually employ highly optimized off-line learning methods. In such systems, a new classifier can only be added if the whole recognition system is redesigned. For many applications that is not a practical approach. To be open for new users and applications, we propose an extensible recognition system with a modular structure. We will show that such an approach can produce almost the same accuracy compared to a system that has been generally trained (only 2 percentage points lower). Our modular classifier system allows the addition of new classifier modules. These modules use Recurrent Fuzzy Inference Systems (RFIS) as mapping functions, that not only deliver a classification, but also an uncertainty value describing the reliability of the classification. Based on the uncertainty value we are able to boost recognition rates. A genetic algorithm search enables the modular combination.

Adaptive content recommendation for mobile users: Ordering recommendations using a hierarchical context model with granularity

Retrieving timely and relevant information on-site is an important task for mobile users. A context-aware system can understand a users information needs and thus select contents according to relevance. We propose a context-dependent search engine that represents user context in a knowledge-based context model, implemented in a hierarchical structure with granularity information. Search results are ordered based on semantic relevance computed as similarity between the current context and tags of search results. Compared against baseline algorithms, the proposed approach enhances precision by 22% and pooled recall by 17%. The use of size-based granularity to compute similarity makes the approach more robust against changes in the context model in comparison to graph-based methods, facilitating import of existing knowledge repositories and end-user defined vocabularies (folksonomies). The reasoning engine being light-weight, privacy protection is ensured, as all user information is processed locally on the users phone without requiring communication with an external server.

A Geometry for Places: Representing Extension and Extended Objects

The article presents a qualitative region-based approach to the representation of extension. A geometry of incidence and ordering is taken as a basis to characterize the concept of extension founded on the congruence of certain regions (called places) which have equal extension into all directions. The notion of extension of regions is derived from the sizes of places—not from the distance between points as in classical geometry—and represented by size intervals. A geometric specification of granular or scale-specific spatial contexts and of the local extension of a region is then derived. Extension relative to a spatial context is used to formally specify conditions under which object regions can be classified e.g. as punctual, linear, or planar in the context.