Showing papers presented at "Location- and Context-Awareness in 2009"

Location Diversity: Enhanced Privacy Protection in Location Based Services

Abstract: Location-based Services are emerging as popular applications in pervasive computing. Spatial k -anonymity is used in Location-based Services to protect privacy, by hiding the association of a specific query with a specific user. Unfortunately, this approach fails in many practical cases such as: (i) personalized services, where the user identity is required, or (ii) applications involving groups of users (e.g., employees of the same company); in this case, associating a query to any member of the group, violates privacy. In this paper, we introduce the concept of Location Diversity , which solves the above-mentioned problems. Location Diversity improves Spatial k -anonymity by ensuring that each query can be associated with at least *** different semantic locations (e.g., school, shop, hospital, etc). We present an attack model that maps each observed query to a linear equation involving semantic locations, and we show that a necessary condition to preserve privacy is the existence of infinite solutions in the resulting system of linear equations. Based on this observation, we develop algorithms that generate groups of semantic locations, which preserve privacy and minimize the expected query processing and communication cost. The experimental evaluation demonstrates that our approach reduces significantly the privacy threats, while incurring minimal overhead.

Activity Recognition from Sparsely Labeled Data Using Multi-Instance Learning

Abstract: Activity recognition has attracted increasing attention in recent years due to its potential to enable a number of compelling context-aware applications. As most approaches rely on supervised learning methods, obtaining substantial amounts of labeled data is often an important bottle-neck for these approaches. In this paper, we present and explore a novel method for activity recognition from sparsely labeled data. The method is based on multi-instance learning allowing to significantly reduce the required level of supervision. In particular we propose several novel extensions of multi-instance learning to support different annotation strategies. The validity of the approach is demonstrated on two public datasets for three different labeling scenarios.

Bluetooth Tracking without Discoverability

Abstract: Outdoor location-based services are now prevalent due to advances in mobile technology and GPS. Indoors, however, even coarse location remains unavailable. Bluetooth has been identified as a potential location technology that mobile consumer devices already support, easing deployment and maintenance. However, Bluetooth tracking systems to date have relied on the Bluetooth inquiry mode to constantly scan for devices. This process is very slow and can be a security and privacy risk. In this paper we investigate an alternative: connection-based tracking. This permits tracking of a previously identified handset within a field of fixed base stations. Proximity is determined by creating and monitoring low-level Bluetooth connections that do not require authorisation. We investigate the properties of the low-level connections both theoretically and in practice, and show how to construct a building-wide tracking system based on this technique. We conclude that the technique is a viable alternative to inquiry-based Bluetooth tracking.

Daily Routine Recognition through Activity Spotting

Abstract: This paper explores the possibility of using low-level activity spotting for daily routine recognition. Using occurrence statistics of low-level activities and simple classifiers based on their statistics allows to train a discriminative classifier for daily routine activities such as working and commuting. Using a recently published data set we find that the number of required low-level activities is surprisingly low, thus, enabling efficient algorithms for daily routine recognition through low-level activity spotting. More specifically we employ the JointBoosting-framework using low-level activity spotters as weak classiers. By using certain low-level activities as support, we achieve an overall recall rate of over 90% and precision rate of over 88%. Tuning down the weak classifiers using only 2.61% of the original data still yields recall and precision rates of 80% and 83%.

Using Context Annotated Mobility Profiles to Recruit Data Collectors in Participatory Sensing

Abstract: Mobile phones and accompanying network layers provide a platform to capture and share location, image, and acoustic data. This substrate enables participatory sensing: coordinated data gathering by individuals and communities to explore the world around them. Realizing such widespread and participatory sensing poses difficult challenges. In this paper, we discuss one particular challenge: creating a recruitment service to enable sensing organizers to select well-suited participants. Our approach concentrates on finding participants based on geographic and temporal coverage, as determined by context-annotated mobility profiles that model transportation mode, location, and time. We outline a three-stage recruitment framework designed to be parsimonious so as to limit risk to participants by reducing the location and context information revealed to the system. Finally, we illustrate the utility of the framework, along with corresponding modeling technique for mobility information, by analyzing data from a pilot mobility study consisting of ten users.

Error Estimation for Indoor 802.11 Location Fingerprinting

Abstract: 802.11-based indoor positioning systems have been under research for quite some time now. However, despite the large attention this topic has gained, most of the research focused on the calculation of position estimates. In this paper, we go a step further and investigate how the position error that is inherent to 802.11-based positioning systems can be estimated. Knowing the position error is crucial for many applications that rely on position information: End users could be informed about the estimated position error to avoid frustration in case the system gives faulty position information. Service providers could adapt their delivered services based on the estimated position error to achieve a higher service quality. Finally, system operators could use the information to inspect whether a location system provides satisfactory positioning accuracy throughout the covered area. For position error estimation, we present four novel algorithms that take different features into account. Advantages of the combination of these four algorithms are explored by using a machine-learning approach. We evaluate our proposed algorithms in two different real-world deployments by using real-world data and emulation. The results show that our algorithms work independently of the environment and the positioning algorithm, and with an average precision for estimating the position error of up to 1.45 meters. The algorithms can --- by adjusting parameters --- realize different tradeoffs between underestimating and overestimating errors. Furthermore we comment on the algorithms' space and time complexity.

Improving Location Fingerprinting through Motion Detection and Asynchronous Interval Labeling

Abstract: Wireless signal strength fingerprinting has become an increasingly popular technique for realizing indoor localization systems using existing WiFi infrastructures. However, these systems typically require a time-consuming and costly training phase to build the radio map. Moreover, since radio signals change and fluctuate over time, map maintenance requires continuous re-calibration. We introduce a new concept called "asynchronous interval labeling" that addresses these problems in the context of user-generated place labels. By using an accelerometer to detect whether a device is moving or stationary, the system can continuously and unobtrusively learn from all radio measurements during a stationary period, thus greatly increasing the number of available samples. Movement information also allows the system to improve the user experience by deferring labeling to a later, more suitable moment. Initial experiments with our system show considerable increases in data collected and improvements to inferred location likelihood, with negligible overhead reported by users.

Multi Activity Recognition Based on Bodymodel-Derived Primitives

Abstract: We propose a novel model-based approach to activity recognition using high-level primitives that are derived from a human body model estimated from sensor data. Using short but fixed positions of the hands and turning points of hand movements, a continuous data stream is segmented in short segments of interest. Within these segments, joint boosting enables the automatic discovery of important and distinctive features ranging from motion over posture to location. To demonstrate the feasibility of our approach we present the user-dependent and across-user results of a study with 8 participants. The specific scenario that we study is composed of 20 activities in quality inspection of a car production process.

Position Estimation from UWB Pseudorange and Angle-of-Arrival: A Comparison of Non-linear Regression and Kalman Filtering

Abstract: This paper presents two algorithms, non-linear regression and Kalman filtering, that fuse heterogeneous data (pseudorange and angle-of-arrival) from an ultra-wideband positioning system The performance of both the algorithms is evaluated using real data from two deployments, for both static and dynamic scenarios We also consider the effectiveness of the proposed algorithms for systems with reduced infrastructure (lower deployment density), and for lower-complexity sensing platforms which are only capable of providing either pseudorange or angle-of-arrival

SofTOA: Software Ranging for TOA-Based Positioning of WLAN Terminals

Abstract: TOA-based trilateration constitutes an interesting choice to locate terminals employing WLAN networks. A major limitation of this technique is the requirement for hardware modifications in the WLAN device in order to achieve accurate ranging. This paper presents an approach that overcomes this limitation. It is based on RTT measurements performed through time-stamping the transmission and reception of IEEE 802.11 MAC frames from the WLAN driver's code, employing the CPU clock as time-base. Some statistical processing is needed in order to mitigate the noise of the measurements. This paper presents experimental results obtained with a prototype showing ranging errors of a few meters when applied to estimate distances up to 25 meters in both indoor and outdoor environments.

Directional Beaconing: A Robust WiFi Positioning Method Using Angle-of-Emission Information

Abstract: WiFi-based positioning has been widely used because it does not require any additional sensors for existing WiFi mobile devices. However, positioning accuracy based on radio signal strength is often influenced by noises, reflections, and obstacles. The Time-of-Arrival (TOA) or Angle-of-Arrival (AOA) methods may be used, but both require additional sensing mechanisms and cannot be applied to existing WiFi mobile devices. In this paper, we propose a new WiFi-based positioning method called directional beaconing. This method uses the Angle-of-Emission (AOE) method instead of the AOA. Using this method, access points (APs) emit beacon signals through rotating directional antennas with angle information encoded in beacons. WiFi devices estimate the direction and distance to the AP by receiving and decoding these beacons. This method integrates the advantages of the AOA and signal strength methods without requiring additional sensors. We constructed revolving directional APs and verified positioning accuracy is improved using the proposed method.

LOCK: A Highly Accurate, Easy-to-Use Location-Based Access Control System

Abstract: With proliferation of ubiquitous computing, digital access is facing an increasing risk since unauthorized client located at any place may intrude a local server. Location Based Access Control (LBAC) is a promising solution that tries to protect the client's access within some user-defined secure zones. Although a lot of prior work has focused on LBAC, most of them suffer from coarse judgment resolution problem or considerable manual setting-up efforts. This paper proposes LOCK, a highly accurate, easy-to-use LBAC system, which uses autonomous ultrasound positioning devices and an access control engine to precisely characterize the secure zones and accurately judge the online access authority. Particularly, the ultrasound positioning device provides relative 3D coordinate of the mobile clients. Measurement-Free Calibration (MFC) is proposed to easily calibrate these positioning devices to transform their relative positioning results into an absolute coordinate system. In this coordinate system, secure zones are characterized by a Coherent Secure Zone Fitting (CSZF) method to compensate the disparity between manually measured secure zone and the secure zone seen by the positioning devices. Furthermore, a Round-Trip Judgment (RTJ) algorithm is designed to fast online determine the geographical relationship between the client's position and such secure zones. A prototype of LOCK system was implemented by defining a meeting table as secure zone to control the client's access to a FTP server. Experiment results show that the system can be easily set up and can control the client's access with centimeter level judgment resolution.

Activity Inference through Sequence Alignment

Abstract: Activity inference attempts to identify what a person is doing at a given point in time from a series of observations. Since the 1980s, the task has developed into a fruitful research field and is now considered a key step in the design of many human-centred systems. For activity inference, wearable and mobile devices are unique opportunities to sense a user's context unobtrusively throughout the day. Unfortunately, the limited battery life of these platforms does not always allow continuous activity logging. In this paper, we present a novel technique to fill in gaps in activity logs by exploiting both short- and long-range dependencies in human behaviour. Inference is performed by sequence alignment using scoring parameters learnt from training data in a probabilistic framework. Experiments on the Reality Mining dataset show significant improvements over baseline results even with reduced training and long gaps in data.

A Framework for Creating and Using Maps of Privately Owned Spaces

Abstract: The Facility Map Framework (FMF) is a new system of specifications and software for maps of privately owned spaces --- campus, building, room detail, and shelf or rack. Based on a generic object model, FMF supports map authoring from CAD or other drawings, asset databases and location sensor systems, map access through browsers or applications, and development of application programs. FMF can integrate existing system components to support interoperability and extensibility, and has features to address a number of additional engineering problems in location systems for private spaces. FMF is a work in progress, with many key elements implemented and others currently under design. Several demos are now in place, representing a variety of different scenarios for using indoor maps. This paper gives an overview of the system components and a number of the engineering challenges and solutions; it represents a progress report on this evolving project.

Nonvisual, Distal Tracking of Mobile Remote Agents in Geosocial Interaction

Abstract: With the recent introduction of mass-market mobile phones with location, bearing and acceleration sensing, we are on the cusp of significant progress in location-based interaction, and highly interactive mobile social networking. We propose that such systems must work when subject to typical uncertainties in the sensed or inferred context, such as user location, bearing and motion. In order to examine the feasibility of such a system we describe an experiment with an eyes-free, mobile implementation which allows users to find a target user, engage with them by pointing and tilting actions, then have their attention directed to a specific target. Although weaknesses in the design of the tilt---distance mapping were indicated, encouragingly, users were able to track the target, and engage with the other agent.

An Ultrasonic 3D Positioning System Using a Single Compact Receiver Unit

Abstract: We present a system that can identify positions of objects in three-dimensional space using ultrasonic communications. The proposed system uses the time of flight of ultrasonic waves to calculate the distance between a transmitter and a microphone. An innovative method called the Phase Accordance Method is used for accurate detection of the arrival time of the ultrasonic signal for distance measurements. The main benefit of the proposed system is that it can locate positions of static and moving objects in a three-dimensional space by using only one compact receiver unit mounting three ultrasonic microphones; this will reduce the users' deployment tasks. Experimental results prove that the system possesses sufficient accuracy levels and stable performance of position measurements in static and dynamic situations, even though the beacon geometry of the system is poor.

Characterizing the Space by Presenting Non-visual Information

Abstract: We are aiming at adding characteristics to the existing space by providing people with non-visual, location-dependent information. By manipulating non-visual information presented to people, we can change implicit partitioning of the space without physically reconstructing it. "Thermotaxis" is a system that gives sensations of cool and warm to users by controlling thermoelectric devices wirelessly. In this system, the space is characterized as being cool or warm. Users experience the difference in temperatures while they walk in the space. Preliminary analysis shows that people stay close in the area of a comfortable temperature.