Jane Yung-jen Hsu

Context-aware taxi demand hotspots prediction

In an urban area, the demand for taxis is not always matched up with the supply. This paper proposes mining historical data to predict demand distributions with respect to contexts of time, weather, and taxi location. The four-step process consists of data filtering, clustering, semantic annotation, and hotness calculation. The results of three clustering algorithms are compared and demonstrated in a web mash-up application to show that context-aware demand prediction can help improve the management of taxi fleets.

The diet-aware dining table: observing dietary behaviors over a tabletop surface

We are what we eat. Our everyday food choices affect our long-term and short-term health. In the traditional health care, professionals assess and weigh each individuals dietary intake using intensive labor at high cost. In this paper, we design and implement a diet-aware dining table that can track what and how much we eat. To enable automated food tracking, the dining table is augmented with two layers of weighing and RFID sensor surfaces. We devise a weight-RFID matching algorithm to detect and distinguish how people eat. To validate our diet-aware dining table, we have performed experiments, including live dining scenarios (afternoon tea and Chinese-style dinner), multiple dining participants, and concurrent activities chosen randomly. Our experimental results have shown encouraging recognition accuracy, around 80%. We believe monitoring the dietary behaviors of individuals potentially contribute to diet-aware healthcare.

KissKissBan: a competitive human computation game for image annotation

In this paper, we propose a competitive human computation game, KissKissBan (KKB), for image annotation. KKB is different from other human computation games since it integrates both collaborative and competitive elements in the game design. In a KKB game, one player, the blocker, competes with the other two collaborative players, the couples; while the couples try to find consensual descriptions about an image, the blockers mission is to prevent the couples from reaching consensus. Because of its design, KKB possesses two nice properties over the traditional human computation game. First, since the blocker is encouraged to stop the couples from reaching consensual descriptions, he will try to detect and prevent coalition between the couples; therefore, these efforts naturally form a player-level cheating-proof mechanism. Second, to evade the restrictions set by the blocker, the couples would endeavor to bring up a more diverse set of image annotations. Experiments hosted on Amazon Mechanical Turk and a gameplay survey involving 17 participants have shown that KKB is a fun and efficient game for collecting diverse image annotations.

Collaborative localization: enhancing WiFi-based position estimation with neighborhood links in clusters

Location-aware services can benefit from accurate and reliable indoor location tracking. The widespread adoption of 802.11x wireless LAN as the network infrastructure creates the opportunity to deploy WiFi-based location services with few additional hardware costs. While recent research has demonstrated adequate performance, localization error increases significantly in crowded and dynamic situations due to electromagnetic interferences. This paper proposes collaborative localization as an approach to enhance position estimation by leveraging more accurate location information from nearby neighbors within the same cluster. The current implementation utilizes ZigBee radio as the neighbor-detection sensor. This paper introduces the basic model and algorithm for collaborative localization. We also report experiments to evaluate its performance under a variety of clustering scenarios. Our results have shown 28.2-56% accuracy improvement over the baseline system Ekahau, a commercial WiFi localization system.

Touching the void: direct-touch interaction for intangible displays

In this paper, we explore the challenges in applying and investigate methodologies to improve direct-touch interaction on intangible displays. Direct-touch interaction simplifies object manipulation, because it combines the input and display into a single integrated interface. While traditional tangible display-based direct-touch technology is commonplace, similar direct-touch interaction within an intangible display paradigm presents many challenges. Given the lack of tactile feedback, direct-touch interaction on an intangible display may show poor performance even on the simplest of target acquisition tasks. In order to study this problem, we have created a prototype of an intangible display. In the initial study, we collected user discrepancy data corresponding to the interpretation of 3D location of targets shown on our intangible display. The result showed that participants performed poorly in determining the z-coordinate of the targets and were imprecise in their execution of screen touches within the system. Thirty percent of positioning operations showed errors larger than 30mm from the actual surface. This finding triggered our interest to design a second study, in which we quantified task time in the presence of visual and audio feedback. The pseudo-shadow visual feedback was shown to be helpful both in improving user performance and satisfaction.

Community-based game design: experiments on social games for commonsense data collection

Games with A Purpose have successfully harvested information from web users. However, designing games that encourage sustainable and quality data contribution remains a great challenge. Given that many online communities have enjoyed active participation from a loyal following, this research explores how human computation games may benefit from rich interactions inherent in a community. We experimented by implementing two games for commonsense data collection on the leading social community platforms: the Rapport Game on Facebook and the Virtual Pet Game on PTT. In this paper, we present the choices of interaction mode and goal-oriented user model for building a community-based game. The data quality, collection efficiency, player retention, concept diversity, and game stability of both games are analyzed quantitatively from data collected since August/November 2008. Our findings should provide useful suggestions for designing community-based games in the future.

Fuzzy classification trees for data analysis

Overly generalized predictions are a serious problem in concept classification. In particular, the boundaries among classes are not always clearly defined. For example, there are usually uncertainties in diagnoses based on data from biochemical laboratory examinations. Such uncertainties make the prediction be more difficult than noise-free data. To avoid such problems, the idea of fuzzy classification is proposed. This paper presents the basic definition of fuzzy classification trees along with their construction algorithm. Fuzzy classification trees is a new model that integrates the fuzzy classifiers with decision trees, that can work well in classifying the data with noise. Instead of determining a single class for any given instance, fuzzy classification predicts the degree of possibility for every class.Some empirical results the dataset from UCI Repository are given for comparing FCT and C4.5. Generally speaking, FCT can obtain better results than C4.5.

Accountability monitoring and reasoning in service-oriented architectures

Service-oriented architecture (SOA) provides a powerful paradigm to compose service processes using individual atomic services. When running a service process, SOA needs an efficient and effective mechanism to detect service delivery failures and to identify the individual service(s) that causes the problem. In this research, we study the model of accountability to detect, diagnose, and defuse the real cause of a problem when service errors (such as incorrect result or SLA violation) occur in a service process. Our approach leverages Bayesian networks to identify the most likely problematic services in a process and selectively inspect those services. An evidence channel selection algorithm is designed to specify which services in a service network should be monitored to achieve the best cost-efficiency. We model the channels selection as the classic facilities location problem. We also adopt a continuous knowledge learning process to manage the dynamic nature of SOA. The performance study shows that our proposed accountability mechanism is effective on identifying the root cause of problems and can achieve significant cost savings: with 50% of services’ outputs monitored as evidence, the comprehensive diagnosis correctness can reach 80% after only 20% of services are inspected.

Applying power meters for appliance recognition on the electric panel

Recognition of appliances states is an import building block for making energy-efficiency schemes and providing energy-saving advice and performing automatic control. Several existing approachs use smart outlets or detectors to acquire the information of individual appliance and recognize the operating state. However, such approachs have to install numerous devices if they want to monitor the states of all appliances. This will increase the cost and complexity of installation and maintenance. Therefore, we develop an appliance recognition system which minimizing the scope of deployment. We install smart meters at single-point, distribution board, to measure the power consumption at circuit-level. In addition, to improve the recognition accuracy of our system and detect the state changes in real time, We use dynamic baysian network to take user behavior into account and Bayes filter to perform online inference. Finally, we design several experiments to compare our approach with some commonly used classifiers, such as Naive Bayes, k-Nearest Neighbor (kNN) and Support Vector Machine (SVM). Results shows that our model outperforms these classifiers and the accuracies of all appliances are greater than 92%. Furthermore, we also compare the results of Bayes filter with Viterbi algorithm, which is an offline inference method. The difference in accuracy of every appliance between Bayes filter and Viterbi algorithm is less than 1%.

Playful tray: adopting Ubicomp and persuasive techniques into play-based occupational therapy for reducing poor eating behavior in young children

This study has created the Playful Tray that adopts Ubicomp and persuasive techniques into play-based occupational therapy for reducing poor eating behavior in young children after they reached their self-feeding age. The design of the Playful Tray reinforces active participation of children in the activity of eating by integrating digital play with eating. Results of a pilot user study suggest that the Playful Tray may improve child meal completion time and reduce negative power play interactions between parents and children, resulting in an improved family mealtime experience.