Hypervideo usage scenarios like physiotherapy trainings or instructions for manual tasks make it hard for users to use an input device like a mouse or touch screen on a hand-held device while they are performing an exercise or use both hands to perform a manual task. In this work, we are trying to overcome this issue by providing an alternative input method for hypervideo navigation using speech commands. In a user test, we evaluated two different speech recognition libraries, annyang (in combination with the Web Speech API) and PocketSphinx.js (in combination with the Web Audio API), for their usability to control hypervideo players. Test users spoke 18 words, either in German or English, which were recorded and then processed by both libraries. We found out that annyang shows better recognition results. However, depending on other factors of influence, like the occurrence of background noise (reliability), the availability of an internet connection, or the used browser, PocketSphinx.js may be a better fit.

Hypervideo based physiotherapy trainings bear an opportunity to support patients in continuing their training after being released from a rehabilitation clinic. Many exercises require the patient to sit on the floor or a gymnastic ball, lie on a gymnastics mat, or do the exercises in other postures. Using a laptop or tablet with a stand to show the exercises is more helpful than for example just having some drawings on a leaflet. However, it may lead to incorrect execution of the exercises while maintaining eye contact with the screen or require the user to get up and select the next exercise if the devices is positioned for a better view. A dual screen application, where contents are shown on a TV screen and the flow of the video can be controlled from a mobile second device, allows patients to keep their correct posture and the same time view and select contents. In this paper we propose first studies for user interface designs for such apps. Initial paper prototypes are discussed and refined in two focus groups. The results are then presented to a broader range of users in a survey. Three prototypes for the mobile app and one prototype for the TV are identified for future user tests.

The creation of hypervideos usually requires a lot of planning and is time consuming with respect to media content creation. However, when structure and media are put together to author a hypervideo, it may only require minor changes to make the hypervideo available in other languages or for another user group (like beginners versus experts). However, to make the translation of media and all navigation elements of a hypervideo efficient and manageable, the authoring tool needs a GUI that provides a good overview of elements that can be translated and of missing translations. In this work, we propose screen concepts that help authors to provide different versions (for example language and/or experience level) of a hypervideo. We analyzed different variants of GUI elements and evaluated them in a survey. We draw guidelines from the results that can help with the creation of similar systems in the future.

The confluence of technologies such as telepresence, immersive imaging, model based virtual mirror worlds, mobile live streaming, etc. give rise to a capability for people anywhere to view and connect with present or past events nearly anywhere on earth. This capability properly belongs to a public commons, available as a birthright of all humans, and can been seen as part of an evolutionary transition supporting a global collective mind. We describe examples and elements of this capability, and suggest how they can be better integrated through a tool we call TeleViewer and a framework called WorldViews, which supports easy sharing of views as well as connecting of providers and consumers of views all around the world.

Most current mobile and wearable devices are equipped with inertial measurement units (IMU) that allow the detection of motion gestures, which can be used for interactive applications. A difficult problem to solve, however, is how to separate ambient motion from an actual motion gesture input. In this work, we explore the use of motion gesture data labeled with gesture execution phases for training supervised learning classifiers for gesture segmentation. We believe that using gesture execution phase data can significantly improve the accuracy of gesture segmentation algorithms. We define gesture execution phases as the start, middle and end of each gesture. Since labeling motion gesture data with gesture execution phase information is work intensive, we used crowd workers to perform the labeling. Using this labeled data set, we trained SVM-based classifiers to segment motion gestures from ambient movement of the device t. We describe initial results that indicate that gesture execution phase can be accurately recognized by SVM classifiers. Our main results show that training gesture segmentation classifiers with phase-labeled data substantially increases the accuracy of gesture segmentation: we achieved a gesture segmentation accuracy of 0.89 for simulated online segmentation using a sliding window approach.

As workstyles change to include more dynamic contexts and denser spaces, connected objects and spaces in the workplace can play a bigger role in helping people get their work done, while also helping them navigate the continually blending boundaries between work- and home lives. In this talk, we argue that the workplace is particularly well-suited for realizing the "connected life" by including both company-initiated sensing in the workplace and personal tracking devices introduced by individual workers. We describe some examples of ubiquitous sensing in the workplace, and future opportunities as well as open technical and ethical issues for designing for the connected [work]life.

Search log analysis has become a common practice to gain insights into user search behaviour, it helps gain an understanding of user needs and preferences, as well as how well a system supports such needs. Currently log analysis is typically focused on the low-level user actions, i.e. logged events such as issued queries and clicked results; and often only a selection of such events are logged and analysed. However, the types of logged events may differ widely from interface to interface, making comparison between systems difficult. Further, analysing a selection of events may lead to conclusions out of context— e.g. the statistics of observed query reformulations may be influenced by the existence of a relevance feedback component. Alternatively, in lab studies user activities can be analysed at a higher level, such as search tactics and strategies, abstracted away from detailed interface implementation. However, the required manual codings that map logged events to higher level interpretations prevent this type of analysis from going large scale. In this paper, we propose a new method for analysing search logs by (semi-)automatically identifying user search tactics from logged events, allowing large scale analysis that is comparable across search systems. We validate the efficiency and effectiveness of the proposed tactic identification method using logs of two reference search systems of different natures: a product search system and a video search system. With the identified tactics, we perform a series of novel log analyses in terms of entropy rate of user search tactic sequences, demonstrating how this type of analysis allows comparisons of user search behaviours across systems of different nature and design. This analysis provides insights not achievable with traditional log analysis.

We propose a method for extractive summarization of audiovisual recordings focusing on topic-level segments. We first build a content similarity graph between all segments of all documents in the collection, using word vectors from the transcripts, and then select the most central segments for the summaries. We evaluate the method quantitatively on the AMI Meeting Corpus using gold standard reference summaries and the Rouge metric, and qualitatively on lecture recordings using a novel two-tiered approach with human judges. The results show that our method compares favorably with others in terms of Rouge, and outperforms the baselines for human scores, thus also validating our evaluation protocol.

Many people post about their daily life on social media. These posts may include information about the purchase activity of people, and insights useful to companies can be derived from them: e.g. profile information of a user who mentioned something about their product. As a further advanced analysis, we consider extracting users who are likely to buy a product from the set of users who mentioned that the product is attractive. In this paper, we report our methodology for building a corpus for Twitter user purchase behavior prediction. First, we collected Twitter users who posted a want phrase + product name: e.g. "want a Xperia" as candidate want users, and also candidate bought users in the same way. Then, we asked an annotator to judge whether a candidate user actually bought a product. We also annotated whether tweets randomly sampled from want/bought user timelines are relevant or not to purchase. In this annotation, 58% of want user tweets and 35% of bought user tweets were annotated as relevant. Our data indicate that information embedded in timeline tweets can be used to predict purchase behavior of tweeted products.

The negative effect of lapses during a behavior-change program has been shown to increase the risk of repeated lapses and, ultimately, program abandonment. In this paper, we examine the potential of system-driven lapse management -- supporting users through lapses as part of a behavior-change tool. We first review lessons from domains such as dieting and addiction research and discuss the design space of lapse management. We then explore the value of one approach to lapse management -- the use of "cheat points" as a way to encourage sustained participation. In an online study, we first examine interpretations of progress that was reached through using cheat points. We then present findings from a deployment of lapse management in a two-week field study with 30 participants. Our results demonstrate the potential of this approach to motivate and change users' behavior. We discuss important open questions for the design of future technology-mediated behavior change programs.

Taking breaks from work is an essential and universal practice. In this paper, we extend current research on productivity in the workplace to consider the break habits of knowledge workers and explore opportunities of break logging for personal informatics. We report on three studies. Through a survey of 147 U.S.-based knowledge workers, we investigate what activities respondents consider to be breaks from work, and offer an understanding of the benefit workers desire when they take breaks. We then present results from a two-week in-situ diary study with 28 participants in the U.S. who logged 800 breaks, offering insights into the effect of work breaks on productivity. We finally explore the space of information visualization of work breaks and productivity in a third study. We conclude with a discussion of implications for break recommendation systems, availability and interuptibility research, and the quantified workplace.

We describe a novel thermal haptic output device, ThermoTouch, that provides a grid of thermal pixels. Unlike previous devices which mainly use Peltier elements for thermal output, ThermoTouch uses liquid cooling and electro-resistive heating to output thermal feedback at arbitrary grid locations. We describe the design of the prototype, highlight advantages and disadvantages of the technique and briefly discuss future improvements and research applications.

Silicon Valley is home to many of the world’s largest technology corporations, as well as thousands of small startups. Despite the development of other high-tech economic centers throughout the US and around the world, Silicon Valley continues to be a leading hub for high-tech innovation and development, in part because most of its companies and universities are within 20 miles of each other. Given the high concentration of multimedia researchers in Silicon Valley, and the high demand for information exchange, I was able to work with a team of researchers from various companies and organizations to start the Bay Area Multimedia Forum (BAMMF) series back in November 2013.

With modern technologies, it is possible to create annotated interactive non-linear videos (a form of hypervideo) for the Web. These videos have a non-linear structure of linked scenes to which additional information (other media like images, text, audio, or additional videos) can be added. A variety of user interactions - like in- and between-scene navigation or zooming into additional information - are possible in players for this type of video. Like linear video, quality of experience (QoE) with annotated hypervideo experiences is tied to the temporal consistency of the video stream at the client end - its flow. Despite its interactive complexity, users expect this type of video experience to flow as seamlessly as simple linear video. However, the added hypermedia elements bog playback engines down. Download and cache management systems address the flow issue, but their effectiveness is tied to numerous questions respecting user requirements, computational strategy, and evaluative metrics. In this work, we a) define QoE metrics, b) examine structural and behavioral patterns of interactive annotated non-linear video, c) propose download and cache management algorithms and strategies, d) describe the implementation of an evaluative simulation framework, and e) present the algorithm test results.

We present a method for profiling businesses at specific locations that is based on mining information from social media. The method matches geo-tagged tweets from Twitter against venues from Foursquare to identify the specific business mentioned in a tweet. By linking geo-coordinates to places, the tweets associated with a business, such as a store, can then be used to profile that business. From these venue-located tweets, we create sentiment profiles for each of the stores in a chain. We present the results as heat maps showing how sentiment differs across stores in the same chain and how some chains have more positive sentiment than other chains. We also estimate social group size from photos and create profiles of social group size for businesses. Sample heat maps of these results illustrate how the average social group size can vary across businesses.

We describe methods for analyzing and visualizing document metadata to provide insights about collaborations over time. We investigate the use of Latent Dirichlet Allocation (LDA) based topic modeling to compute areas of interest on which people collaborate. The topics are represented in a node-link force directed graph by persistent fixed nodes laid out with multidimensional scaling (MDS), and the people by transient movable nodes. The topics are also analyzed to detect bursts to highlight "hot" topics during a time interval. As the user manipulates a time interval slider, the people nodes and links are dynamically updated. We evaluate the results of LDA topic modeling for the visualization by comparing topic keywords against the submitted keywords from the InfoVis 2004 Contest, and we found that the additional terms provided by LDA-based keyword sets result in improved similarity between a topic keyword set and the documents in a corpus. We extended the InfoVis dataset from 8 to 20 years and collected publication metadata from our lab over a period of 21 years, and created interactive visualizations for exploring these larger datasets.

The use of videoconferencing in the workplace has been steadily growing. While multitasking during video conferencing is often necessary, it is also viewed as impolite and sometimes unacceptable. One potential contributor to negative attitudes towards such multitasking is the disrupted sense of eye contact that occurs when an individual shifts their gaze away to another screen, for example, in a dual-monitor setup, common in office settings. We present a system to improve a sense of eye contact over videoconferencing in dual-monitor setups. Our system uses computer vision and desktop activity detection to dynamically choose the camera with the best view of a user's face. We describe two alternative implementations of our system (RGB-only, and a combination of RGB and RGB-D cameras). We then describe results from an online experiment that shows the potential of our approach to significantly improve perceptions of a person's politeness and engagement in the meeting.

This paper presents a detailed examination of factors that affect perceptions of, and attitudes towards multitasking in dyadic video conferencing. We first report findings from interviews with 15 professional users of videoconferencing. We then report results from a controlled online experiment with 397 participants based in the United States. Our results show that the technology used for multitasking has a significant effect on others' assumptions of what secondary activity the multitasker is likely engaged in, and that this assumed activity in turn affects evaluations of politeness and appropriateness. We also describe how different layouts of the video conferencing UI may lead to better or worse ratings of engagement and in turn ratings of polite or impolite behavior. We then propose a model that captures our results and use the model to discuss implications for behavior and for the design of video communication tools.

We present MixMeetWear, a smartwatch application that allows users to maintain awareness of the audio and visual content of a meeting while completing other tasks. Users of the system can listen to the audio of a meeting and also view, zoom, and pan webcam and shared content keyframes of other meeting participants' live streams in real time. Users can also provide input to the meeting via speech-to-text or predefined responses. A study showed that the system is useful for peripheral awareness of some meetings.

Remote meetings are messy. There are an ever-increasing number of support tools available, and, as past work has shown, people will tend to select a subset of those tools to satisfy their own institutional, social, and personal needs. While video tools make it relatively easy to have conversations at a distance, they are less adapted to sharing and archiving multimedia content. In this paper we take a deeper look at how sharing multimedia content occurs before, during, and after distributed meetings. Our findings shed light on the decisions and rationales people use to select from the vast set of tools available to them to prepare for, conduct, and reconcile the results of a remote meeting.

In recent years, there has been an explosion of services that lever- age location to provide users novel and engaging experiences. However, many applications fail to realize their full potential because of limitations in current location technologies. Current frameworks work well outdoors but fare poorly indoors. In this paper we present LoCo, a new framework that can provide highly accurate room-level indoor location. LoCo does not require users to carry specialized location hardware—it uses radios that are present in most contemporary devices and, combined with a boosting classification technique, provides a significant runtime performance improvement. We provide experiments that show the combined radio technique can achieve accuracy that improves on current state-of-the-art Wi-Fi only techniques. LoCo is designed to be easily deployed within an environment and readily leveraged by application developers. We believe LoCo’s high accuracy and accessibility can drive a new wave of location-driven applications and services.

Image localization is important for marketing and recommendation of local business; however, the level of granularity is still a critical issue. Given a consumer photo and its rough GPS information, we are interested in extracting the fine-grained location information (i.e. business venues) of the image. To this end, we propose a novel framework for business venue recognition. The framework mainly contains three parts. First, business aware visual concept discovery: we mine a set of concepts that are useful for business venue recognition based on three guidelines including business-awareness, visually detectable, and discriminative power. Second, business-aware concept detection by convolutional neural networks (BA-CNN): we pro- pose a new network architecture that can extract semantic concept features from input image. Third, multimodal business venue recognition: we extend visually detected concepts to multimodal feature representations that allow a test image to be associated with business reviews and images from social media for business venue recognition. The experiments results show the visual concepts detected by BA-CNN can achieve up to 22.5% relative improvement for business venue recognition compared to the state-of-the-art convolutional neural network features. Experiments also show that by leveraging multimodal information from social media we can further boost the performance, especially in the case when the database images belonging to each business venue are scarce.

Hypervideos yield to different challenges in the area of navigation due to their underlying graph structure. Especially when used on tablets or by older people, a lack of clarity may lead to confusion and rejection of this type of medium. To avoid confusion, the hypervideo can be extended with a well known table of contents, which needs to be created separately by the authors due to an underlying graph structure. In this work, we present an extended presentation of a table of contents for hypervideos on mobile devices. The design was tested in a real world medical training scenario with the target group of people older than 45 which is the main target group of these applications. This user group is a particular challenge since they sometimes have limited experience in the use of mobile devices and physical deficiencies with growing age. Our user interface was designed in three steps. The findings of an expert group and a survey were used to create two different prototypical versions of the display, which were then tested against each other in a user test. This test revealed that a divided view is desired. The table of contents in an easy-to-touch version should be on the left side and previews of scenes should be on the right side of the view. These findings were implemented in the existing SIVA HTML5 open source player and tested with a second group of users. This test only lead to minor changes in the GUI.

Indoor localization is challenging in terms of both the accuracy and possible using scenarios. In this paper, we introduce the design and implementation of a toy car localization and navigation system, which demonstrates that a projected light based localization technique allows multiple devices to know and exchange their fine-grained location information in an indoor environment. The projected light consists of a sequence of gray code images which assigns each pixel in the projection area a unique gray code so as to distinguish their coordination. The light sensors installed on the toy car and the potential “passenger” receive the light stream from the projected light stream, based on which their locations are computed. The toy car then utilizes A* algorithm to plan the route based on its own location, orientation, the target’s location and the map of available “roads”. The fast speed of localization enables the toy car to adjust its own orientation while “driving” and keep itself on “roads”. The toy car system demonstrates that the localization technique can power other applications that require fine-grained location information of multiple objects simultaneously.

In this paper, we analyze the association between a social media user's photo content and their interests. Visual content of photos is analyzed using state-of-the-art deep learning based automatic concept recognition. An aggregate visual concept signature is thereby computed for each user. User tags manually applied to their photos are also used to construct a tf-idf based signature per user. We also obtain social groups that users join to represent their social interests. In an effort to compare the visual-based versus tag-based user profiles with social interests, we compare corresponding similarity matrices with a reference similarity matrix based on users' group memberships. A random baseline is also included that groups users by random sampling while preserving the actual group sizes. A difference metric is proposed and it is shown that the combination of visual and text features better approximates the group-based similarity matrix than either modality individually. We also validate the visual analysis against the reference inter-user similarity using the Spearman rank correlation coefficient. Finally we cluster users by their visual signatures and rank clusters using a cluster uniqueness criteria.