Page 73 - Undergraduate Research Opportunities Program Proceedings 2017-18

School of Engineering
Department of Computer Science & Engineering

views of the answer; Users are most interested in topics about the country they reside in and topics their alma mater are
well-known for; Answerers of a certain topic are also interested in topics with similar concepts; Accuracy and the
number of association rules among topics both decreases as the minimum support in the FP-growth algorithm increases.

Augmented Reality on Wearable Devices

Supervisor: HUI Pan / CSE
Student: YEUNG Ngo Yan / COGBM Course: UROP1100, Summer

This report investigates on the possibility of a pressure-based text input system derived from traditional text on 9 keys
(T9) predictive text keyboard. First, the report briefly introduces the constraints of migrating existing keyboards into
devices with a limited key factor, such as smartwatches plus a brief introduction to some of the force-assisted input
method so far. Next, the report will include the implementation of the keyboard, including two preliminary test designs
set up to determine the sensibility of users towards exerting different numbers of levels of pressure. The keyboard also
comes with a haptic feedback mechanics to provide feedback referring to the level of pressure exerted. The keyboard
can potentially be used in devices with limited form factors or incorporated with smart glasses or other display devices
to provide an eyes-off text entry system. The system is implemented on an iPhone 6s with force touch functionality and
Apple Taptic engine as a prior study. Finally, possible future improvements to the keyboard will also be discussed at the
end of the report.

Supervisor: HUI Pan / CSE
Student: YU Jiuzhi / COMP Course: UROP1100, Spring

The main feature of augmented reality is to mix the digital information with the concrete objects in real life, in order to
provide the user with an immersive experience. With the help of the advanced computer vision techniques (object
recognition, etc.), it is far easier to detect the objects and information needed by the user. The next step is to collect the
correct data about the scene and display them accurately and succinctly. The primary focus of this UROP project is about
finding a feasible solution for providing a stable information service for augmented reality applications, providing the
information from the scene is correctly detected.

Augmented Reality Technology for Visually Impaired

Supervisor: HUI Pan / CSE
Student: ZHANG Daofu / SENG Course: UROP1100, Summer

New technology advances such as wearables, computer vision can also lead to new shopping experiences and analytics.
However, many current approaches lack of precision (e.g., wireless) or threats customer’s privacy (e.g., face recognition,
smartphone tracking). In this report, we propose an innovative system to track shopper’s gaze and provide better
insights about customers’ behavior in physical stores. EyeShopper is a gaze estimator ubiquitous system which includes
privacy awareness (i.e., face recognition) and works on video surveillance footage. We believe that the fine-grained
information collected by our system can be the key to success for retail stores. We define customer’s gaze as a two-
parameter function based on head pose: yaw, and pitch. In this report we will focus on back head detection algorithms
so the system will be able to not only estimate clients’ gazes but track them. So this gaze estimations can be provided to
retailers as another level of granularity (i.e., what customers look at and for how long) to enhance their store analytics
with augmented features such as most viewed products.

63