Showing papers presented at "International Symposium on Wearable Computers in 2013"

Don't mind me touching my wrist: a case study of interacting with on-body technology in public

Abstract: Wearable technology, specifically e-textiles, offers the potential for interacting with electronic devices in a whole new manner. However, some may find the operation of a system that employs non-traditional on-body interactions uncomfortable to perform in a public setting, impacting how readily a new form of mobile technology may be received. Thus, it is important for interaction designers to take into consideration the implications of on-body gesture interactions when designing wearable interfaces. In this study, we explore the third-party perceptions of a user's interactions with a wearable e-textile interface. This two-prong evaluation examines the societal perceptions of a user interacting with the textile interface at different on-body locations, as well as the observer's attitudes toward on-body controller placement. We performed the study in the United States and South Korea to gain cultural insights into the perceptions of on-body technology usage.

Preference, context and communities: a multi-faceted approach to predicting smartphone app usage patterns

Abstract: Reliable smartphone app prediction can strongly benefit both users and phone system performance alike. However, real-world smartphone app usage behavior is a complex phenomena driven by a number of competing factors. In this pa- per, we develop an app usage prediction model that leverages three key everyday factors that affect app usage decisions -- (1) intrinsic user app preferences and user historical patterns; (2) user activities and the environment as observed through sensor-based contextual signals; and, (3) the shared aggregate patterns of app behavior that appear in various user communities. While rapid progress has been made recently in smartphone app prediction, existing prediction models tend to focus on only one of these factors. We evaluate a multi-faceted approach to prediction using (1) a 3-week 35-user field trial, along with (2) analysis of app usage logs of 4,606 smartphone users worldwide. We find our app usage model can not only produce more robust app predictions than conventional techniques, but it can also enable significant smartphone system optimizations.

On preserving statistical characteristics of accelerometry data using their empirical cumulative distribution

Abstract: The majority of activity recognition systems in wearable computing rely on a set of statistical measures, such as means and moments, extracted from short frames of continuous sensor measurements to perform recognition. These features implicitly quantify the distribution of data observed in each frame. However, feature selection remains challenging and labour intensive, rendering a more generic method to quantify distributions in accelerometer data much desired. In this paper we present the ECDF representation, a novel approach to preserve characteristics of arbitrary distributions for feature extraction, which is particularly suitable for embedded applications. In extensive experiments on six publicly available datasets we demonstrate that it outperforms common approaches to feature extraction across a wide variety of tasks.

I know what you are reading: recognition of document types using mobile eye tracking

Abstract: Reading is a ubiquitous activity that many people even perform in transit, such as while on the bus or while walking. Tracking reading enables us to gain more insights about expertise level and potential knowledge of users -- towards a reading log tracking and improve knowledge acquisition. As a first step towards this vision, in this work we investigate whether different document types can be automatically detected from visual behaviour recorded using a mobile eye tracker. We present an initial recognition approach that com- bines special purpose eye movement features as well as machine learning for document type detection. We evaluate our approach in a user study with eight participants and five Japanese document types and achieve a recognition performance of 74% using user-independent training.

FIDO - facilitating interactions for dogs with occupations: wearable dog-activated interfaces

Abstract: Working dogs have improved the lives of thousands of people. However, communication between human and canine partners is currently limited. The main goal of the FIDO project is to research fundamental aspects of wearable technologies to support communication between working dogs and their handlers. In this pilot study, the FIDO team investigated on-body interfaces for assistance dogs in the form of wearable technology integrated into assistance dog vests. We created four different sensors that dogs could activate (based on biting, tugging, and nose gestures) and tested them on-body with three assistance-trained dogs. We were able to demonstrate that it is possible to create wearable sensors that dogs can reliably activate on command.

Personalized mobile physical activity recognition

Abstract: Personalization of activity recognition has become a topic of interest recently. This paper presents a novel concept, using a set of classifiers as general model, and retraining only the weight of the classifiers with new labeled data from a previously unknown subject. Experiments with different methods based on this concept show that it is a valid approach for personalization. An important benefit of the proposed concept is its low computational cost compared to other approaches, making it also feasible for mobile applications. Moreover, more advanced classifiers (e.g. boosted decision trees) can be combined with the new concept, to achieve good performance even on complex classification tasks. Finally, a new algorithm is introduced based on the proposed concept, which outperforms existing methods, thus further increasing the performance of personalized applications.

Sensor-embedded teeth for oral activity recognition

Abstract: This paper presents the design and implementation of a wearable oral sensory system that recognizes human oral activities, such as chewing, drinking, speaking, and coughing. We conducted an evaluation of this oral sensory system in a laboratory experiment involving 8 participants. The results show 93.8% oral activity recognition accuracy when using a person-dependent classifier and 59.8% accuracy when using a person-independent classifier.

Improved actionSLAM for long-term indoor tracking with wearable motion sensors

Abstract: We present an indoor tracking system based on two wearable inertial measurement units for tracking in home and workplace environments. It applies simultaneous localization and mapping with user actions as landmarks, themselves recognized by the wearable sensors. The approach is thus fully wearable and no pre-deployment effort is required. We identify weaknesses of past approaches and address them by introducing heading drift compensation, stance detection adaptation, and ellipse landmarks. Furthermore, we present an environment-independent parameter set that allows for robust tracking in daily-life scenarios. We assess the method on a dataset with five participants in different home and office environments, totaling 8.7h of daily routines and 2500m of travelled distance. This dataset is publicly released. The main outcome is that our algorithm converges 87% of the time to an accurate approximation of the ground truth map (0.52m mean landmark positioning error) in scenarios where previous approaches fail.

Detecting bends and fabric folds using stitched sensors

Abstract: In this paper we describe a novel method for detecting bends and folds in fabric structures. Bending and folding can be used to detect human joint angles directly, or to detect possible errors in the signals of other joint-movement sensors due to fabric folding. Detection is achieved through measuring changes in the resistance of a complex stitch, formed by an industrial coverstitch machine using an un-insulated conductive yarn, on the surface of the fabric. We evaluate self-intersecting folds which cause short-circuits in the sensor, creating a quasi-binary resistance response, and non-contact bends, which deform the stitch structure and result in a more linear response. Folds and bends created by human movement were measured on the dorsal and lateral knee of both a robotic mannequin and a human. Preliminary results are promising. Both dorsal and lateral stitches showed repeatable characteristics during testing on a mechanical mannequin and a human.

Eyeglass-based hands-free videophone

Abstract: We propose an eyeglass-based videophone that enables the wearer to make a video call without holding a phone (that is to say hands-free) in the mobile environment. The glasses have 4 (or 6) fish-eye cameras to widely capture the face of the wearer and the images are fused to yield 1 frontal face image. The face image is also combined with the background image captured by a rear-mounted camera; the result is a self-portrait image without holding any camera device at arm's length. Simulations confirm that 4 fish-eye cameras with 250-degree field of view (or 6 cameras with 180-degree field of view) can cover 83% of the frontal face. We fabricate a 6 camera prototype, and confirm the possibility of generating the self-portrait image. This system suits not only hands-free videophones but also other applications like visual life logging and augmented reality use.

Retrofitting smartphones to be used as particulate matter dosimeters

Abstract: This work discusses ways of measuring particulate matter with mobile devices. Solutions using a dedicated sensor device are presented along with a novel method of retrofitting a sensor to a camera phone without need for electrical modifications. Instead, the flash and camera of the phone are used as light source and receptor of an optical dust sensor respectively. Experiments to evaluate the accuracy are presented.

Confidence-based multiclass AdaBoost for physical activity monitoring

Abstract: Physical activity monitoring has recently become an important topic in wearable computing, motivated by e.g. healthcare applications. However, new benchmark results show that the difficulty of the complex classification problems exceeds the potential of existing classifiers. Therefore, this paper proposes the ConfAdaBoost.M1 algorithm. The proposed algorithm is a variant of the AdaBoost.M1 that incorporates well established ideas for confidence based boosting. The method is compared to the most commonly used boosting methods using benchmark datasets from the UCI machine learning repository and it is also evaluated on an activity recognition and an intensity estimation problem, including a large number of physical activities from the recently released PAMAP2 dataset. The presented results indicate that the proposed ConfAdaBoost.M1 algorithm significantly improves the classification performance on most of the evaluated datasets, especially for larger and more complex classification tasks.

Ultrasound-based movement sensing, gesture-, and context-recognition

Abstract: We propose an activity and context recognition method where the user carries a neck-worn receiver comprising a microphone, and small speakers on his wrists that generate ultrasounds. The system recognizes gestures on the basis of the volume of the received sound and the Doppler effect. The former indicates the distance between the neck and wrists, and the later indicates the speed of motions. Thus, our approach substitutes the wired or wireless communication typically required in body area motion sensing networks by ultrasounds. Our system also recognizes the place where the user is in and the people who are near the user by ID signals generated from speakers placed in rooms and on people. The strength of the approach is that, for offline recognition, a simple audio recorder can be used for the receiver. We evaluate the approach in one scenario on nine gestures/activities with 10 users. Evaluation results confirmed that when there was no environmental sound generated from other people, the recognition rate was 87% on average. When there was environmental sound generated from other people, we compare approach ultrasound-based recognition which uses only the feature value of ultrasound against standard approach, which uses feature value of ultrasound and environmental sound. Results for the proposed approach are 65%, for the standard approach are 57%.

A hybrid unsupervised/supervised model for group activity recognition

Abstract: The new method proposed here recognizes activities performed by a group of users (e.g., attending a meeting, playing sports, and participating in a party) by using sensor data obtained from the users. Note that such group activities (GAs) have characteristics that differ from those of single user activities. For example, the number of users who participate in a GA is different for each activity. The number of meeting participants, for instance, may sometimes be different for each meeting. Also, a user may play different roles (e.g., `moderator' and `presenter' roles) in meetings on different days. We introduce the notion of role into our GA recognition model and try to capture the intrinsic characteristics of GAs with a hybrid unsupervised/supervised approach.

Sensing group proximity dynamics of firefighting teams using smartphones

Abstract: Firefighters work in dangerous and unfamiliar situations under a high degree of time pressure and thus team work is of utmost importance. Relying on trained automatisms, firefighters coordinate their actions implicitly by observing the actions of their team members. To support training instructors with objective mission data, we aim to automatically detect when a firefighter is in-sight with other firefighters and to visualize the proximity dynamics of firefighting missions. In our approach, we equip firefighters with smartphones and use the built-in ANT protocol, a low-power communication radio, to measure proximity to other firefighters. In a second step, we cluster the proximity data to detect moving sub-groups. To evaluate our method, we recorded proximity data of 16 professional firefighting teams performing a real-life training scenario. We manually labeled six training sessions, involving 51 firefighters, to obtain 79 minutes of ground truth data. On average, our algorithm assigns each group member to the correct ground truth cluster with 80% accuracy. Considering height information derived from atmospheric pressure signals increases group assignment accuracy to 95%.

Wearable partner agent with anthropomorphic physical contact with awareness of user's clothing and posture

Abstract: In this paper, we introduce a wearable partner agent, that makes physical contacts corresponding to the user's clothing, posture, and detected contexts. Physical contacts are generated by combining haptic stimuli and anthropomorphic motions of the agent. The agent performs two types of the behaviors: a) it notifies the user of a message by patting the user's arm and b) it generates emotional expression by strongly enfolding the user's arm. Our experimental results demonstrated that haptic communication from the agent increases the intelligibility of the agent's messages and familiar impressions of the agent.

Can i wash it?: the effect of washing conductive materials usedin making textile based wearable electronic interfaces.

Abstract: We explore the wash-ability of conductive materials used in creating traces and touch sensors in wearable electronic textiles. We perform a wash test measuring change in resistivity after each of 10 cycles of washing for conductive traces constructed using two types of conductive thread, conductive ink, and combinations of thread and ink.

Conductive rubber electrodes for earphone-based eye gesture input interface

Abstract: An eartip made of conductive rubber that also realizes bio-potential electrodes is proposed for a daily-use earphone-based eye gesture input interface. Several prototypes, each with three electrodes to capture Electrooculogram (EOG), are implemented on earphones and examined. Experiments with one subject over a 10 day period reveal that all prototypes capture EOG similarly but they differ as regards stability of the baseline and the presence of motion artifacts. Another experiment conducted on a simple eye-controlled application with six subjects shows that the proposed prototype minimizes motion artifacts and offers good performance. We conclude that conductive rubber with mixed Ag filler is the most suitable setup for daily-use.

ThermOn: thermo-musical interface for an enhanced emotional experience

Abstract: This report proposes a thermal media system, ThermOn, which enables users to feel dynamic hot and cold sensations on their body corresponding to the sound of music. Thermal sense plays a significant role in the human recognition of environments and influences human emotions. By employing thermal sense in the music experience, which also greatly affects human emotions, we have successfully created a new medium with an unprecedented emotional experience. With ThermOn, a user feels enhanced excitement and comfort, among other responses. For the initial prototype, headphone-type interfaces were implemented using a Peltier device, which allows users to feel thermal stimuli on their ears. Along with the hardware, a thermal-stimulation model that takes into consideration the characteristics of human thermal perception was designed. The prototype device was verified using two methods: the psychophysical method, which measures the skin potential response and the psychometric method using a Likert-scale questionnaire and open-ended interviews. The experimental results suggest that ThermOn (a) changes the impression of music, (b) provides comfortable feelings, and (c) alters the listener's ability to concentrate on music in the case of a rock song. Moreover, these effects were shown to change based on the methods with which thermal stimuli were added to music (such as temporal correspondence) and on the type of stimuli (warming or cooling). From these results, we have concluded that the ThermOn system has the potential to enhance the emotional experience when listening to music.

Wearable computing: through the looking glass

Abstract: Google's Glass has captured the world's imagination, with new articles speculating on it almost every day. Yet, why would consumers want a wearable computer in their everyday lives? For the past 20 years, my teams have been creating living laboratories to discover the most compelling reasons. In the process, we have investigated how to create interfaces for technology which are designed to be "there when you need it, gone when you don't." This talk will attempt to articulate the most valuable lessons we have learned, including some design principles for creating "microinteractions" to fit a user's lifestyle.

Activity monitoring in daily life as an outcome measure for surgical pain relief intervention using smartphones

Abstract: We investigate the potential of a smartphone to measure a patient's change in physical activity before and after a surgical pain relief intervention. We show feasibility for our smartphone system providing physical activity from acceleration, barometer and location data to measure the intervention's outcome. In a single-case study, we monitored a pain patient carrying the smartphone before and after a surgical intervention over 26 days. Results indicate significant changes before and after intervention, particularly in physical activity in the home environment.

3D from looking: using wearable gaze tracking for hands-free and feedback-free object modelling

Abstract: This paper presents a method for estimating the 3D shape of an object being observed using wearable gaze tracking. Starting from a sparse environment map generated by a simultaneous localization and mapping algorithm (SLAM), we use the gaze direction positioned in 3D to extract the model of the object under observation. By letting the user look at the object of interest, and without any feedback, the method determines 3D point-of-regards by back-projecting the user's gaze rays into the map. The 3D point-of-regards are then used as seed points for segmenting the object from captured images and the calculated silhouettes are used to estimate the 3D shape of the object. We explore methods to remove outlier gaze points that result from the user saccading to non object points and methods for reducing the error in the shape estimation. Being able to exploit gaze information in this way, enables the user of wearable gaze trackers to be able to do things as complex as object modelling in a hands-free and even feedback-free manner.

Reversible contacting of smart textiles with adhesive bonded magnets

Abstract: The aim of this study was to develop a reversible electrical contacting through adhesive bonded neodymium magnets. To implement this, suitable magnets and adhesives are chosen by defined requirements and conductive bonds between textile and magnet are optimized. For the latter, three different bonds are produced and tested in terms of achievable conductivity and mechanical strength. It is shown that gold-coated neodymium magnets are most appropriate for such a contact. The reproducible electrical resistances are low with sufficient mechanical strength.

Scaled monocular SLAM for walking people

Abstract: In this paper we present a full-scaled real-time monocular SLAM using only a wearable camera. Assuming that the person is walking, the perception of the head oscillatory motion in the initial visual odometry estimate allows for the computation of a dynamic scale factor for static windows of N camera poses. Improving on this method we introduce a consistency test to detect non-walking situations and propose a sliding window approach to reduce the delay in the update of the scaled trajectory. We evaluate our approach experimentally on a unscaled visual odometry estimate obtained with a wearable camera along a path of 886 m. The results show a significant improvement respect to the initial unscaled estimate with a mean relative error of 0.91% over the total trajectory length.

Pattern resistors: exploring resistive motifs as components for e-embroidery

Abstract: E-textile practitioners have improvised innovatively with existing off-the shelf electronics to make them textile-compatible However, there is a need to further the development of soft materials or parts that could replace regular electronics in a circuit As a starting point, we look at the possibility of creating a repository of specific motifs with different resistance values that can be easily incorporated into e-embroidery projects and used instead of normal resistors The paper describes our larger objective and gives an overview of the first experiment done to compare the resistance values of a simple pattern embroidered multiple times with conductive yarn to observe its behavior and reliability

An underwater wearable computer for two way human-dolphin communication experimentation

Abstract: Research in dolphin cognition and communication in the wild is still a challenging task for marine biologists. Most problems arise from the uncontrolled nature of field studies and the challenges of building suitable underwater research equipment. We present a novel underwater wearable computer enabling researchers to engage in an audio-based interaction between humans and dolphins. The design requirements are based on a research protocol developed by a team of marine biologists associated with the Wild Dolphin Project.

Detecting strumming action while playing guitar

Abstract: In this paper we describe a wristwatch-like device using a 3-axis gyro sensor to determine how a player is strumming the guitar. The device was worn on the right-handed player's right hand to evaluate the strumming action, which is important to play the guitar musically in terms of the timing and the strength of notes. With a newly developed calculation algorithm to specify the timing and the strength of the motion when the guitar string(s) were strummed, beginners and experienced players were clearly distinguished without hearing the sounds. The beginners as well as intermediate-level players showed a fairly large variation of the maximum angular velocity around the upper arm for each strum. Since the developed system reports the evaluation results with a graphical display as well as sound effects in real time, the players may improve their strumming action without playing back the performance.

Reducing user intervention in incremental activityrecognition for assistive technologies

Abstract: Activity recognition has recently gained a lot of interest and there already exist several methods to detect human activites based on wearable sensors. Most of the existing methods rely on a database of labelled activities that is used to train an offline activity recognition system. This paper presents an approach to build an online activity recognition system that do not require any a priori labelled data. The system incrementally learns activities by actively querying the user for labels. To choose when the user should be queried, we compare a method based on random sampling and another that uses a Growing Neural Gas (GNG). The use of GNG helps reducing the number of user queries by 20% to 30%.

Prior knowledge of human activities from social data

Abstract: We explore the feasibility of utilizing large, crowd-generated online repositories to construct prior knowledge models for high-level activity recognition. Towards this, we mine the popular location-based social network, Foursquare, for geo-tagged activity reports. Although unstructured and noisy, we are able to extract, categorize and geographically map people's activities, thereby answering the question: what activities are possible where? Through Foursquare text only, we obtain a testing accuracy of 59.2% with 10 activity categories; using additional contextual cues such as venue semantics, we obtain an increased accuracy of 67.4%. By mapping prior odds of activities via geographical coordinates, we directly benefit activity recognition systems built on geo-aware mobile phones.

Driving low-power wearable systems with an adaptively-controlled foot-strike scavenging platform

Abstract: We explore the use of Dielectric Elastomer (DE) micro-generators as a means to scavenge energy from foot-strikes and power wearable systems. While they exhibit large energy densities, DEs must be closely controlled to maximize the energy they transduce. Towards this end, we propose a DE micro-generator array configuration that enhances transduction efficiency, and the use of foot pressure sensors to realize accurate control of the individual DEs. Statistical techniques are applied to customize performance for a user's gait and enable energy-optimized adaptive online control of the system. Simulations based on experimentally collected foot pressure datasets, empirical characterization of DE mechanical behavior and a detailed model of DE electrical behavior show that the proposed system can achieve between 45 and 66mJ per stride.