Kai Kunze | Osaka Prefecture University (original) (raw)

Papers by Kai Kunze

Abstract This thesis investigates, how placement variations of electronic devices influence the p... more Abstract This thesis investigates, how placement variations of electronic devices influence the possibility of using sensors integrated in those devices for context recognition. The vast majority of context recognition research assumes well defined, fixed sensor locations. Although this might be acceptable for some application domains (eg in an industrial setting), users, in general, will have a hard time coping with these limitations.

Pervasive …, Jan 1, 2009

We describe a systematic, quantitative study of the benefits using context recognition (specifica... more We describe a systematic, quantitative study of the benefits using context recognition (specifically task tracking) for a wearable maintenance assistance system. A key objective of the work is to do the evaluation in an environment that is as close as possible to a real world setting. To this end, we use actual maintenance tasks on a complex piece of machinery at an industrial site. Subjects for our study are active Zeiss technicians who have an average of 10 years job experience.

Apparatus and methods are provided for determining a person's compass orientation from data gener... more Apparatus and methods are provided for determining a person's compass orientation from data generated by one or more sensors embedded within a portable device transported by the person. Such sensors may include an accelerometer, a GPS (Global Positioning System) receiver, a magnetometer (eg, a compass), a barometer, etc.

Abstract In a Wizard-of-Oz experiment we investigate to what degree automatic activity recognitio... more Abstract In a Wizard-of-Oz experiment we investigate to what degree automatic activity recognition could support the use of prioritized lists for nursing documentation on hand held mobile devices. The study involved 15 nurses, 60 patients records and over 250 documented processes at a geriatric care ward. Based on time effort, interaction complexity, error rate and subjective system perception our wizard of oz study shows that activity recognition is a key factor in the usability and acceptance of the system.

We investigate the use of magnetic field disturbances as features for motion based, wearable acti... more We investigate the use of magnetic field disturbances as features for motion based, wearable activity recognition systems. Such disturbances are mostly caused by large metallic objects and electrical appliances, both of which are often involved in human activities. We propose to detect them by subtracting angular velocity values computed from the changes in the magnetic field vector from gyroscope signals.

Abstract General representation, abstraction and exchange definitions are crucial for dynamically... more Abstract General representation, abstraction and exchange definitions are crucial for dynamically configurable context recognition. However, to evaluate potential definitions, suitable standard datasets are needed. This paper presents our effort to create and maintain large scale, multimodal standard datasets for context recognition research. We ourselves used these datasets in previous research to deal with placement effects and presented low-level sensor abstractions in motion based on-body sensing.

Proceedings of the 12th …, Jan 1, 2010

The demo will present a tool chain for recording, monitoring, labeling, and manipulation of compl... more The demo will present a tool chain for recording, monitoring, labeling, and manipulation of complex multimodal data sets for activity recognition. The tool chain is comprehensive (going from logging, through labeling, monitoring to post processing and managing the data), integrated (with all tools being able to cooperate on joint data sets), and build around comfortable graphical user interfaces.

… Journal of Sensors, …, Jan 1, 2011

Opportunistic sensing can be used to obtain data from sensors that just happen to be present in t... more Opportunistic sensing can be used to obtain data from sensors that just happen to be present in the user's surroundings. By harnessing these opportunistic sensor configurations to infer activity or context, ambient intelligence environments become more robust, have improved user comfort thanks to reduced requirements on body-worn sensor deployment and they are not limited to a predefined and restricted location, defined by sensors specifically deployed for an application.

… Computers, 2008. ISWC …, Jan 1, 2008

We demonstrate how modulated magnetic field technology that is well established in high precision... more We demonstrate how modulated magnetic field technology that is well established in high precision, stationary motion tracking systems can be adapted to wearable activity recognition. To this end we describe the design and implementation of a cheap (components cost about 20 Euro for the transmitter and 15 Euro for the receiver), low power (17mA for the transmitter and 40mA for the receiver), and easily wearable (the main size constraint are the coils which are about 25mm 3 ) system for tracking the relative position and orientation of body parts. We evaluate our system on two recognition tasks. On a set of 6 subtle nutrition related gestures it achieves 99.25% recognition rate compared to 94.1% for a XSens inertial device ( operated calibrated, euler angle mode). On the recognition of 8 Tai Chi moves it reaches 94 % compared to 86% of an accelerometer. Combining our sensor with the accelerometer leads to 100% correct recognition (as compared to 90% when combining the accelerometer with a gyro).

UbiComp 2007: Ubiquitous Computing, Jan 1, 2007

We describe a novel method for symbolic location discovery of simple objects. The method requires... more We describe a novel method for symbolic location discovery of simple objects. The method requires no infrastructure and relies on simple sensors routinely used in sensor nodes and smart objects (acceleration, sound). It uses vibration and short, narrow frequency 'beeps' to sample the response of the environment to mechanical stimuli. The method works for specific locations such as 'on the couch', 'in the desk drawer' as well as for location classes such as 'closed wood compartment' or 'open iron surface'. In the latter case, it is capable of generalizing the classification to locations the object has not seen during training. We present the results of an experimental study with a total of over 1200 measurements from 35 specific locations (taken from 3 different rooms) and 12 abstract location classes. It includes such similar locations as the inner and outer pocket of a jacket and a table and shelf made of the same wood. Nonetheless on locations from a single room (16 in the largest one) we achieve a recognition rate of up to 96 %. It goes down to 81 % if all 35 locations are taken together, however the correct location is in the 3 top picks of the system 94 % of the times.

Proceedings of the …, Jan 1, 2010

We investigate if and how magnetic sensors can be used to replace gyroscopes in wearable activity... more We investigate if and how magnetic sensors can be used to replace gyroscopes in wearable activity recognition. The work is motivated by (1) sensor configurations typically found in smart phones where magnetic sensors are used to complement GPS position with orientation and (2) the fact that gyroscopes are an important source of information for activity recognition. We propose a method to compute angular velocity from 3D magnetic sensor data and discuss its fundamental limitations. We present an elaborate evaluation of the accuracy on 5 previously published data set with a total of nearly 20 hours of data from 15 users with activities ranging from bicycle repair, through homemaking to gym exercise.

… , 2009. ISWC'09. …, Jan 1, 2009

We present a method to infer the orientation of mobile device carried in a pocket from the accele... more We present a method to infer the orientation of mobile device carried in a pocket from the acceleration signal acquired when the user is walking. Whereas previous work has shown how to determine the the orientation in the vertical plane (angle towards earth gravity), we demonstrate how to compute the orientation within the horizontal plane.

Wearable Computers, 2007 11th IEEE …, Jan 1, 2007

This paper is part of an effort to facilitate wearable activity recognition using dynamically cha... more This paper is part of an effort to facilitate wearable activity recognition using dynamically changing sets of sensors integrated in everyday appliances such as phones, PDAs, watches, headsets etc. A key issue that such systems have to address is the position of the devices on the body. In general each devices can be in a number of different locations (e.g. headset on the head or in on of many pockets). At the same time most activity recognition algorithms require fixed, known sensor positions.

… and Games, 2007. …, Jan 1, 2007

We envision to add context awareness and ambient intelligence to edutainment and computer gaming ... more We envision to add context awareness and ambient intelligence to edutainment and computer gaming applications in general. This requires mixed-reality setups and ever-higher levels of immersive human-computer interaction. Here, we focus on the automatic recognition of natural human hand gestures recorded by inexpensive, wearable motion sensors. To study the feasibility of our approach, we chose an educational parking game with 3-D graphics that employs motion sensors and hand gestures as its sole game controls. Our implementation prototype is based on Java-3D for the graphics display and on our own CRN Toolbox for sensor integration. It shows very promising results in practice regarding game appeal, player satisfaction, extensibility, ease of interfacing to the sensors, and -last but not least -sufficient accuracy of the real-time gesture recognition to allow for smooth game control. An initial quantitative performance evaluation confirms these notions and provides further support for our setup.

Ambient …, Jan 1, 2003

The paper describes initial results in an ongoing project aimed at providing and analyzing standa... more The paper describes initial results in an ongoing project aimed at providing and analyzing standardized, representative data sets for typical context recognition tasks. Such data sets can be used to develop user independent feature sets and recognition algorithms. In addition, we aim to establish standard benchmark data sets that can be used for quantitative comparisons of different recognition methodologies. Such benchmark data sets are commonly used speech and image recognition, but so far none are available for general context recognition tasks. We outline the experimental considerations and procedures used to record the data in a controlled manner, observing strict experimental standards. We then discuss preliminary results obtained with common features on a well-understood scenario with 8 subjects. The discussion shows that even for a small sample like this variations between subjects are substantial, thus underscoring the need for large representative data sets.

This paper describes a concept called LifeNet for using an ad-hoc sensor network providing relati... more This paper describes a concept called LifeNet for using an ad-hoc sensor network providing relative positioning and a wearable system to support firefighters at indoor-navigation under impaired visibility. We describe the concept and a first implementation.

… (IFAWC), 2006 3rd …, Jan 1, 2006

Inexpensive wearable sensors are well-suited for the automatic recognition of many activities occ... more Inexpensive wearable sensors are well-suited for the automatic recognition of many activities occuring in everyday life. But what about fast and involved movements such as those occuring in athletic sports? We tackle this question by studying the feasibility of using bodyworn gyroscopes and acceleration sensors to recognize Tai Chi movements. To this end, we conducted an initial experiment with eight sensors each affixed to four different persons who repeatedly performed three distinct Tai Chi movements. The resulting data confirm that standard thresholding and pattern-matching techniques should suffice to automate the analysis and recognition of the movements. Moreover, the data also seem to allow for distinguishing between certain levels of expertise and quality in executing the movements.

World of Wireless, …, Jan 1, 2009

Opportunistic sensing allows to efficiently collect information about the physical world and the ... more Opportunistic sensing allows to efficiently collect information about the physical world and the persons behaving in it. This may mainstream human context and activity recognition in wearable and pervasive computing by removing requirements for a specific deployed infrastructure. In this paper we introduce the newly started European research project OPPORTUNITY within which we develop mobile opportunistic activity and context recognition systems. We outline the project's objective, the approach we follow along opportunistic sensing, data processing and interpretation, and autonomous adaptation and evolution to environmental and user changes, and we outline preliminary results.

Architecture of Computing …, Jan 1, 2006

We present a GUI-based C++ toolbox that allows for building distributed, multi-modal context reco... more We present a GUI-based C++ toolbox that allows for building distributed, multi-modal context recognition systems by plugging together reusable, parameterizable components. The goals of the toolbox are to simplify the steps from prototypes to online implementations on low-power mobile devices, facilitate portability between platforms and foster easy adaptation and extensibility. The main features of the toolbox we focus on here are a set of parameterizable algorithms including different filters, feature computations and classifiers, a runtime environment that supports complex synchronous and asynchronous data flows, encapsulation of hardware-specific aspects including sensors and data types (e.g., int vs. float), and the ability to outsource parts of the computation to remote devices. In addition, components are provided for group-wise, event-based sensor synchronization and data labeling. We describe the architecture of the toolbox and illustrate its functionality on two case studies that are part of the downloadable distribution.

… . 12th Annual IEEE …, Jan 1, 2004

To fully take advantage of high-speed networks while freeing CPU cycles for application processin... more To fully take advantage of high-speed networks while freeing CPU cycles for application processing, the industry is proposing new techniques relying on an extended role of the network interface card such as TCP Offload Engine and Remote Direct Memory Access. This paper presents an experimental study aimed at collecting the performance data needed to assess these techniques. This work is based on the emulation of an advanced network interface card plugged on the I/O bus. In the experimental setting, a processor of a partitioned SMP machine is dedicated to network processing. Achieving a faithful emulation of a network interface card is one of the main concerns and it is guiding the design of the Offload Engine software. This setting has the advantage of being flexible so that many different offload scenarios can be evaluated. Preliminary throughput results of an emulated TCP Offload Engine demonstrate a large benefit. The emulated TCP Offload Engine indeed yields 600 to 900% improvement while still relying on memory copies at the kernel boundary.