AUTHORS: Tossaporn Santad, Piyarat Silapasupphakornwong, Worawat Choensawat, and Kingkarn Sookhanaphibarn

ABSTRACT: We proposed an abandoned-baggage detection system that the baggage was left in public places for security reasons, i.e., subway stations. The proposed system applied the YOLO deep learning model for object detection, and presented a GUI for supporting a parameter setting. With this GUI, the detection system will be invariant to lighting and camera position.

AUTHORS: Worawat Choensawat, Sachie Takahashi, Minako Nakamura, Woong Choi, and Kozaburo Hachimura

ABSTRACT: We have long been conducting research on the description and reproduction of body motion based on Labanotation, and developed a system called LabanEditor. Labanotation uses a symbolic description and it is said that the notation can even describe motions of the fine each finger of a dancer. However, since, in this case, the resulting score staff will become extremely complicated, only the basic description has usually been used. On the other hand, there are styles of motion particular to traditional dances, and if we restrict ourselves to handling these dances, the basic notation scheme is insufficient to describe these motions. Based on the above, we investigated a method of describing and reproducing CG animation of highly-stylized traditional Noh plays using plain Labanotation. This has been realized by preparing motion template files which represent specific motions in Noh plays. We can handle any stylized dance motions other than Noh by preparing separate motion template files. In addition, we made improvements in the functionality of LabanEditor as follows: importing VRML models of stages, background sceneries and characters for dancers which correspond to the various types of dance, and controlling the quality of movements. The evaluation showed that both the system and the method satisfied our initial requirements.

AUTHORS: Termpetch Sookhanaphibarn, Kingkarn Sookhanaphibarn

ABSTRACT: This paper presented the learning experience supplementary program using video clips to teach table tennis for enhancing learning outcome of undergraduate students. The video clips were taken by students during their practices. The study type is an experimental research, and the purpose of this study was to investigate the patterns of video clips that are required to learn the table tennis course and compare the effectiveness of two table tennis instruction methods-conventional method and using video clips. The findings from the study indicate, the most important table tennis skill is bounce the ball, knock board are basic skills and fore hand, back hand rally with partner are higher skills before developing advanced skills. The necessary format of table tennis video clips skills includes: 1) bounce the ball fore and back switching, 2) knock board rally, 3) fore hand rally with partner and 4) back hand rally with partner.

AUTHORS: Kingkarn Sookhanaphibarn, Welaiporn Phukongchai, Tossaporn Santad, Worawat Choensawat

ABSTRACT: This paper presented a game-based rehabilitation of the upper limb after stroke. We designed and developed a game for supporting stroke patients to have an exercise their arms, and the game had functions for recording their playing and showing a performance report. The performance report can infer the progress of bilateral uppper-limb rehabilitation and use for comparing among patient cases. This is because the game used a Kinect device to detect the arm movements in aspect of precision and speed.

AUTHORS: Nao Shikanai, Worawat Choensawat, and Kozaburo Hachimura

ABSTRACT: The purpose of this research is to show movement characteristics of bodies during dance performance. We examined bodily characteristics of dancers when moving and conforming to other dancers. Specifically, we analyzed paired dancer movements through motion capture and compared all body-part characteristics based on data derived from cross-correlation analysis and variance-covariance matrix using an exponential map. We found that the correlation coefficients of speed between the head, shoulders, and knees of dancers were significantly high. The values of variance-covariance through the exponential map based on velocity were positively high between the shoulders vertically, and the hips and knees in a front-back direction. The results indicated that not only the legs but also the shoulders of both dancers move at a similarly fast rate when they coordinate and synchronize with one another.

AUTHORS: Pujana Paliyawan, Worawat Choensawat, and Ruck Thawonmas

ABSTRACT: This paper presents a novel approach for motion segmentation by using strategies of splitting and remerging. The presented approach, Mossar, hybridizes two existing ones to obtain their potential advantages while covering weaknesses: (1) velocity-based, one of the most widely used approaches that has fairly low accuracy but provides computational simplicity and (2) graph-based, a state-of-the-art approach that provides outstanding accuracy, yet bears high computational complexity and a burden in setting of thresholds. An initial set of key frames is generated by a velocity-based splitting process and then fed into a graph-based remerging process for refinement. We present mechanisms that improve key-frames capturing in the velocity-based approach as well as details on how the graph-based approach is modified and later applied to remerging. The proposed approach also allows users to interactively add or reduce the number of key frames to control segmentation hierarchy without the need to change threshold values and rerun segmentation, as usually done in existing approaches. Our experimental results show that the presented hybrid approach, compared to both velocity-based and graph-based, demonstrates superior performance in terms of accuracy and in comparison to graph-based, our approach has not only less complexity but also a lesser number of thresholds, the values of which can be much more simply determined.

AUTHORS: Komal Naranga, Kingkarn Sookhanaphibarna, and Prasong Praneetpolgrangb

ABSTRACT: This research presents a model for malware detection on mobile operating system based on analyzing the operation codes. The research processes are as follows: (1) collecting of both malicious and benign codes on android operating system, (2) extracting features based on the distribution of n-grams frequency where the parameter n = 3 is used, and (3) constructing a model for classification the malicious codes using the extracted features for both malicious and benign codes. In the experiment, 304 malicious codes and 553 benign codes were using to construct the model. The experiment shows that the model achieved more than 85.52% accuracy. For the sensitivity and specificity, the model achieved 71.26% and 90.52%, respectively.

AUTHORS: Shefali Sachdeva, Non Member, IAENG, Romuald Jolivot, Non Member, IAENG, and Worawat Choensawat, Non Member, IAENG

ABSTRACT: Due to the remarkable increased popularity of android based smart phones, the rise in malware targeting these devices is clearly visible. In this paper, a machine learning based technique is proposed to classify android applications in three classes based on the confidence level defined as safe, suspicious and highly suspicious. In this paper, 35 features are extracted and selected from Mobile Security Framework based on penetration testing. A set of experiments has been conducted on the scale of 14,073 android applications which includes 10,000 android applications downloaded from apk-dl.com, 3041 malware and 1032 benign applications. In order to compare the accuracy of the classification model, a ground truth of the confidence level is created by using VirusTotal. The proposed method can detect and classify android applications into three confidence levels with 81.8% accuracy. Experiment for binary classification, classify as being malware or benign,has yielded 93.63% accuracy.

AUTHORS: Kittipat Savetratanakaree, Member, IAENG Kingkarn Sookhanaphibarn, Sarun Intakosum and Ruck Thawonmas

ABSTRACT: In this paper, we propose a new approach to over-sample new minority-class instances along the borderline using the Euclidean distance in the feature space to improve support vector machine (SVM) performance in imbalanced data environments. SVM has been an outstandingly successful classifier in a wide variety of applications where balanced class data distribution is assumed. However, SVM is ineffective when coping with imbalanced datasets whereby the majority-class instances far outnumber the minority-class instances. Our new approach, called Borderline Over-sampling in the Feature Space, can deal with imbalanced data to effectively recognize new minority-class instances for better classification with SVM. The results of our class prediction experiments using the proposed approach demonstrate better performance than the existing SMOTE, Borderline-SMOTE and borderline over-sampling methods in terms of the g-mean and F-measure.

AUTHORS: Worawat Choensawat and Piruna Polsiri

ABSTRACT: This paper introduces the use of Adaptive Neuro-Fuzzy Inference Systems (ANFIS) into the area of finance for Thai firms. This study started with collecting financial data from 82 finance companies and 15 commercial banks operating in the period 1992-1997, before the East Asian economic crisis occurred. Financial data on failed and non-failed firms were then examined to develop fuzzy rules based on CAMEL variables. ANFIS is applied to the area of finance for Thai firms for constructing failure prediction models. These models show that prediction accuracy is greater than 90 percent for one to five years prior to failure, indicating the robustness of models over time. In experiments, models yield more accurate forecasting than a logistic model that has been used in the area of finance for Thai firms. The purpose of this study is to present that models using ANFIS are better suited for financial data sets with high nonlinearity than a logistic model.

AUTHORS: Kittipat Savetratanakaree, Kingkarn Sookhanaphibarn, Sarun Intakosum, Ruck Thawonmas, and Kuan-Ta Chen

ABSTRACT: Most business models of online game companies usually depend on sale of virtual items and the monthly subscription fees. The prediction of player departure could increase revenues by giving special promotions out to the players who are expected to unsubscribe or quit playing the game. This paper proposes a departure prediction approach by using a new feature called “SLKdays” and a game revisitation. The feature "SLKdays" is defined as "Staytime", which is the time each player spending in an online game, of the last k days, and a game revisitation is the playing frequency in the last month to predict the next month subscription. We explore our new feature "SLKdays" to determine the optimal number of k for the departure prediction. With our proposed feature, the accuracy of the departure prediction is high, 91.92%, and the precision and recall rate are 98.22% and 84.51%.

AUTHORS: Akekachai Sacfung, Kingkarn Sookhanaphibarn, and Worawat Choensawat

ABSTRACT: Virtual worlds are successfully used for training a personal skill in an emergency situation. In this paper, we present a serious game for fire evacuation training in any public places like office buildings, conventional centers, hotels, and universities. The goal of fire safety evacuation plan is to minimize the damage, i.e. the number of injured people and casualties. In this paper, we built a virtual environment where imitates a classroom building in Bangkok University. We described a design process and development approach for achieving the objectives of evacuation training. The user evaluation was also conducted, showing that the prototype can be used effectively for training a fire drill at other public places.