Hyun-Tae Jeong

Live Book: A mixed reality book using a projection system

In this paper, we propose a mixed reality book service, called Live Book, which provides new user experience by directly augmenting digital contents on a printed book using a small projector. For this service, a stand-type system was developed, which recognizes each page of the printed book and then automatically projects digital contents corresponding to that page. Main feature of the Live Book is its ability to allow user to interact with the system using a bare finger.

A Distributed Wearable System Based on Multimodal Fusion

Wearable computer can be worn anytime with the support of unrestricted communications and variety of services which provides maximum capability of information use. Key challenges in developing such wearable computers are level of comfort that users do not feel what they wear, easy and intuitive user interface and power management technique. This paper suggests a wearable system that consists of a wristwatch-type gesture recognition device and a personal mobile gateway that functional input/output modules can be freely plugged in and taken out. We describe our techniques implemented during our wearable system development: 1) multimodal fusion engine that recognizes voice and gesture simultaneously, 2) power management technique and 3) gesture recognition engine. Finally, we evaluate the performance of our multimodal fusion engine, and show the power consumption measurement data of our system built with the power management technique.

Hand segmentation and fingertip detection for interfacing of stereo vision-based smart glasses

In this paper, we propose a hand segmentation and fingertip detection method for interacting with projected digital contents in the binocular smart glasses, with a built-in stereo cameras. In order to make robust hand segmentation under the various environmental conditions, we used the adaptive skin color model which is created by fusion of depth and color information. Also, we have defined the finger range to reduce time complexity and to find correct fingertip in the process of fingertip detection. Since it provides a stable hand gesture interface, it can be applied to various applications for smart glasses without the restriction of the mobility and the inconvenience.

Estimating the Number of Clusters with Database for Texture Segmentation Using Gabor Filter

This paper addresses a novel solution of the problem of image segmentation by its texture using Gabor filter. Texture segmentation has been worked well by using Gabor filter, but there still is a problem; the number of clusters. There are several studies about estimating number of clusters with statistical approaches such as gap statistic. However, there are some problems to apply those methods to texture segmentation in terms of accuracy and time complexity. To overcome these limits, this paper proposes novel method to estimate optimal number of clusters for texture segmentation by using training dataset and several assumptions which are appropriate for image segmentation. We evaluate the proposed method on dataset consists of texture image and limit possible number of clusters from 2 to 5. And we also evaluate the proposed method by real image contains various texture such as rock stratum.

Automatic topic-based CF recommendation method considering subject similarity

This paper proposes an automatic Topic-based CF Recommendation (TCFR) method for constructing a combined media with multiple unit media. To predict the preference rating score of the retrieved unit media, the proposed method uses subject similarity of the existing media combination that has been included the unit media and the target media. The subject similarity is determined by the similarity of topic-vectors that represent a set of subject topics of the media. It distinguishes a group of similar media combinations and other one using the similarity of topic vector. The method predicts the final rating score by applying weight value for the rating score that is predicted in the each group based on the typical CF (Collaborative Filtering). In the experiment of this paper, it demonstrates that the proposed context-aware recommendation method considering the subject similarity improves the performance of recommendation more than does the typical user-based CF method.

Tactile Representation of Image in a Plane Actuator.

Our study investigated the effect of vibration elements on the recognition of texture in a planar type actuator. A planar type actuator is highly advantageous in naturally conveying the texture of a graphic object on a large-sized touch screen display to a user, which can considerably improve the users experience when interacting with such screens. In order to investigate the effect of frequency and amplitude elements of vibration on texture representation, we analyzed the texture image quantitatively and conducted a user experiment to find the vibration elements that best matched a texture. Experimental results show that controlling frequency is more effective in expressing texture than controlling amplitude, but in a certain frequency band, different texture can be represented by changing the amplitude only.

Force and displacement analysis of a haptic touchscreen

This paper presents the result of an analysis of the operating characteristics of a haptic touchscreen. The haptic touchscreen was fabricated by mounting an electrostatic parallel plate actuator between a touch panel and a display module. The operating characteristics of the haptic touchscreen were examined in terms of the force and displacement in the vertical direction. The results show that even in the case of the minimum force (near 10 mN), the haptic touchscreen is able to produce sufficient force to perceive the vibration for a wide range of frequencies, considering the threshold at which the user detects the vibration through the fingertips.

Performance comparison of tap gestures on small-screen touch devices

This paper describes a new type of tapping touch gesture called different-finger tap. We classified four tapping gestures as expanding tap gestures and performed a user experiment to investigate their tapping speed and error count during button touch interaction together with the different-finger tap gesture. The results showed that the different-finger tap gesture led to faster tapping time than the long-tap gesture and a lower error count than that seen with the double-tap gesture, although users were required to perform complex mental activities to use the different-finger tap gesture because it was unfamiliar. Participants reported that the different-finger tap gesture was the most suitable for tasks that require the tapping of small buttons correctly and quickly, such as text entry, when it was used together with the single-tap gesture.

Development of multimodal surface interface for audio and tactile feedback

This paper deals with the development of a surface multimodal interface system for audio and tactile feedback. The multimodal interface system consists of a PVDF film and an ITO glass. We have controlled the applied voltage between the PVDF film and ITO glass for the tactile feedback. Similarly, we have controlled the applied voltage between the transparent electrodes on both sides of PVDF film for audio feedback. This paper shows the performance of the developed surface multimodal interface system.

Description and recognition based on directional motion vector for spatial hand gestures

This paper presents a description method and an adaptive DTW recognition algorithm (a-DTW) based on Directional Motion Vector (DMV) for spatial hand gestures. The a-DTW based on the classic DTW (Dynamic Time Warping) is designed to compare the degree of similarity between hand gesture segments and hand gesture commands that are defined as a sequence of DMVs. This study uses IMU (Inertial Measurement Unit) sensors such as an accelerometer and a gyroscope to separate hand gestures into a sequence of DMVs. The proposed description and recognition method are implemented and evaluated on the MCU of a test platform of wrist-band type. The experiment average recognition rate of a-DTW is more than 95% for 7 kinds of spatial hand gestures, and the recognition process has been completed within less than 20ms. By the experiment results, the proposed a-DTW recognition method is useful to improve the degree of freedom for spatial hand gestures.