Smart Sensing and Context

Activity recognition is an important application of body sensor networks. To this end, accurate segmentation of different episodes in the data stream is a pre-requisite of subsequent pattern classification. Current techniques for this purpose tend to require specific supervised learning, thus limiting their general application to pervasive sensing applications. This paper presents an improved multiple eigenspace segmentation algorithm that addresses the common problem of under-segmentation in episode detection. Results show that the proposed algorithm significantly increases the segmentation accuracy when compared to existing methods.

Because health condition and quality of life are directly influenced by the amount and intensity of daily physical activity, monitoring the level of activity has gained interest in recent years for various medical and wellbeing applications. In this paper we describe our experience with implementing and evaluating physical activity monitoring and stimulation using wireless sensor networks and motion sensors. Our prototype provides feedback on the activity level of users using a simple colored light. We conduct experiments on multiple test subjects, performing multiple normal daily activities. The results from our experiments represent the motivation for and a first step towards robust complex physical activity monitoring with multiple sensors distributed over a person’s body. The results show that using a single sensor on the body is inadequate in certain situations. Results also indicate that feedback provided on a person’s activity level can stimulate the person to do more exercise. Using multiple sensor nodes and sensor modalities per subject would improve the activity estimation performance, provided that the sensor nodes are small and inconspicuous.

Several navigation situations can be imagined where visual cueing is not practical or unfeasible, and where the hands are required exclusively for a certain task. The utilization of the sense of touch, as relatively new notification modality, should provide sufficient possibilites to cope with this issue.

The focus in this research work is on two questions (i) how the distance encoding schemas affects the overall navigation speed (or in more detail to what level the time lag contributes to the navigation precision) and (ii) if, beside the vibro-tactile stimulation, the transmission of the noise generated by the individual vibration elements influences the speed and/or precision of route guiding. To deal with these questions we have defined and conducted three waypoint following experiments with two different tactor activation methods, one without and the other two with the distance encoded in the vibration patterns. Additionally, we did studies where we masked the noise of the vibration elements and compared the results against the general setting where masking was not applied.

Our results shows that notification latency led to an increasing number of walking anomalies and consequently affects the walking precision and time to a high degree. Furthermore, we could not find evidence that multimodal stimulation with both vibration force and vibration “noise” tends to result in an increased system performance compared to the system with unimodal feedback using vibrations only.

Sensor networks generate a compact form of data in order to efficiently use the available power of the device. Query engines require a richer form of this data in order that users can express meaningful and useful queries. This enrichment of the sensor stream leads to an inevitable increase in the volume of data transported across the network. If this data is enriched and stored in a database on the same server and used only for offline queries, matters such as enrichment time and increased data volumes can usually be managed quite efficiently. However, when there is a requirement to query sensor output in real time, the time required to enrich data and increased data volumes lead to slow query response times. In this research, we present the liveSensor system which offers the benefits of an enriched sensor stream, providing a high level query interface to the user. However, it delays the enrichment of sensor data until after query results have been generated and thus, maintains a high level of performance by processing raw sensor streams.

During the development of context aware applications a context management component must traditionally be created. This task requires specialist context lifecycle management expertise and hence can be a significant deterrent to application development. It also removes the developers focus from differentiation of their application to an oft repeated development task. This issue can be addressed by encapsulating the context management lifecycle within a web-service, thus providing applications with a low-overhead alternative to managing their context data. The adoption of a web-based approach maximizes the potential number of interacting applications, including smart spaces, web and mobile applications, due to ease of access and widespread support of web technologies. The contribution of this paper is the development of a lifecycle, based on existing work on enterprise data and context aware lifecycles, which is optimized for web-based context management services (WCXMS) and the provision of a web-service implementation of the lifecycle.

Developments in (wireless) sensor and actuator networks and the capabilities to manufacture low cost and energy efficient networked embedded devices have lead to considerable interest in adding real world sense to the Internet and the Web. Recent work has raised the idea towards combining the Internet of Things (i.e. real world resources) with semantic Web technologies to design future service and applications for the Web. In this paper we focus on the current developments and discussions on designing Semantic Sensor Web, particularly, we advocate the idea of semantic annotation with the existing authoritative data published on the semantic Web. Through illustrative examples, we demonstrate how rule-based reasoning can be performed over the sensor observation and measurement data and linked data to derive additional or approximate knowledge. Furthermore, we discuss the association between sensor data, the semantic Web, and the social Web which enable construction of context-aware applications and services, and contribute to construction of a networked knowledge framework.

Geographical information retrieval (GIR) can benefit from context information to adapt the results to a user’s current situation and personal preferences. In this respect, semantics-based GIR is especially challenging because context information – such as collected from sensors – is often provided through numeric values, which need to be mapped to ontological representations based on nominal symbols. The Web Ontology Language (OWL) lacks mathematical processing capabilities that require free variables, so that even basic comparisons and distance calculations are not possible. Therefore, the context information cannot be interpreted with respect to the task and the current user’s preferences. In this paper, we introduce an approach based on semantic rules that adds these processing capabilities to OWL ontologies. The task of recommending personalized surf spots based on user location and preferences serves as a case study to evaluate the capabilities of semantic rules for context-aware geographical information retrieval. We demonstrate how the Semantic Web Rule Language (SWRL) can be utilized to model user preferences and how execution of the rules successfully retrieves surf spots that match these preferences. While SWRL itself enables free variables, mathematical functions are added via built-ins – external libraries that are dynamically loaded during rule execution. Utilizing the same mechanism, we demonstrate how SWRL built-ins can query the Semantic Sensor Web to enable the consideration of real-time measurements and thus make geographical information retrieval truly context-aware.

Acquisition and dissemination of user and environment context information is critical in development and deployment of context-aware systems. It is fundamental to the success of such systems that they have access to a scaleable, robust and flexible context provisioning framework capable of working across all types of devices and networks. In this paper, we present the design, implementation and experiences of developing a context management system that incorporates these ideas. It is based on a consumer-provider broker model, where providers employ a common context representation format, decoupling various entities involved in the production and consumption of context information. We demonstrate how the idea of independent context providers can aid in end-to-end working of a context management framework. One of the major advantages compared to other approaches is the extendibility of the system. By progressively adding Context Providers to legacy mobile communication systems, new context domains can be added. The system is able to evolve constantly and support a variety of emerging context-aware services and applications.

The European FP7 research project m:Ciudad - a metropolis of ubiquitous services - aims at the empowerment of users to create services on mobile terminals. The project demonstrates various scenarios in which users either act as creator of services or interact with the system to search for services or service construction components. The search and recommendation process in the system facilitates retrieval of related entities and shows the results to the users according to their preference and contextual information. The paper demonstrates an approach to integrate contextual information with other search attributes to enable efficient service retrieval and recommendation in mobile user and application scenarios. The specifics of a mobile environment are taken into consideration and are reflected in the design of the m:Ciudad Search and Recommendation Engine. We discuss how context-awareness and proactivity can be implemented and utilised for mobile services.

A robust activity and context-recognition system must be capable of operating over a long period of time, exploiting new sources of information as they become available and evolving in an autonomous manner, coping with user variability and changes in the number and type of available sensors. In particular, wearable and ambient nodes should be trained lifelong, as new context instances naturally arise, and the labeling of the instances should be carried out ideally with no user intervention. In this paper we show by means of an experiment and simulations that we can indeed achieve lifelong learning and automatic labeling by using Context Cells, an architecture capable of sensing, learning, classifying data and exchanging information.

A major challenge in using multi-modal, distributed sensor systems for activity recognition is to maintain a temporal synchronization between individually recorded data streams. A common approach is to use well defined ‘synchronization actions’ performed by the user to generate, easily identifiable pattern events in all recorded data streams. The events are then used to manually align data streams. This paper proposes an automatic method for this synchronization.

We demonstrate that synchronization actions can be automatically identified and used for stream synchronization across widely different sensors such as acceleration, sound, force, and a motion tracking system. We describe fundamental properties and bounds of our event-based synchronization approach. In particular, we show that the event timing relation is transitive for sensor groups with shared members. We analyzed our synchronization approach in three studies. For a large dataset of 5 users and totally 308 data stream minutes we achieved a synchronization error of 0.3 s for more than 80% of the stream.

In the domain of ubiquitous computing, the ability to identify the occurrence of situations is a core function of being ’context-aware’. Given the uncertain nature of sensor information and inference rules, reasoning techniques that cater for uncertainty hold promise for enabling the inference process. In our work, we apply the Dempster Shafer theory of evidence to infer situation occurrence with minimal use of training data. We describe a set of evidential operations for sensor mass functions using context quality and evidence accumulation for continuous situation detection. We demonstrate how our approach enables situation inference with uncertain information using a case study based on a published smart home activity data set.

This paper builds on previous work by different authors on monitoring the use of household devices through analysis of the power line current. Whereas previous work dealt with detecting which device is being used, we go a step further and analyze how the device is being used. We focus on a kitchen scenario where many different devices are relevant to activity recognition. The paper describes a smart, easy to install sensor that we have built to do the measurements and the algorithms which can for example determine the consistency of the substance in the mixer, how many eggs are being boiled (and if they are soft or hard), what size of coffee has been prepared or whether a cutting machine was used to cut bread or salami. A set of multi user experiments has been performed to validate the algorithms.

Finding solutions for the current period of climate change or “global warming” is possibly the most serious and pressing challenge faced by scientists and the wider community today. Although governments are beginning to act, a community wide approach is needed with a large proportion of individuals engaging to reduce energy consumption that depends on fossil fuels. The Passive House (or Passivhaus) standard is an ultra-low energy standard for building construction and design that aims at dramatically reducing energy consumption in the home. While appropriate for new builds, this standard may be difficult to achieve with existing buildings. In this work, Wireless Sensor Network (WSN) technology is examined as an enabling tool to support rapid progression to improved energy efficiency and increased comfort for existing buildings. As with participatory urban sensing, the home occupant could, in the future, take on the role of scientist; developing an awareness of trouble spots in the house would allow them to target problems thus reducing the need for heating and improving comfort. The paper reports on experiences and findings from several residential and commercial environmental monitoring WSN deployments using a WSN developed from off the shelf components. The sensors deployed measure temperature, relative humidity, CO2 concentration and light. Depending on the size and layout of the space to be monitored, added to the scope of deployment, between 12 and 20 nodes were deployed and the monitoring period was 7-14 days per location. The paper illustrates the value of using WSN technologies as enablers for the amateur eco-home scientist on the path towards reduced energy consumption and increased comfort. It evaluates the suitability of the system for both commercial and residential deployments and shows how large quantities of data can be reduced to meaningful high level information delivered to the user.

Smart cars are promising application domain for ubiquitous computing. Context recognition is important support for a smart car to avoid accidents proactively. Despite many techniques have been developed, we find a lack of complex situation recognition in the smart car environment. This paper presents a novel context recognition approach that is composed of two parts: offline statistic-based situation pattern training and online situation recognition. The training phase is done to learn the statistical relationship between simple context atoms and complex context situations and hence generate the pattern of every single situation. The online recognition phase will recognize the current situation according to its pattern in the running time of a smart car. The implementation of the software and prototype is given to provide the running environment for the approach. Performance evaluation shows that our approach is effective and applicable in a smart car.