Bunthit Watanapa

Human gesture recognition using Kinect camera

In this paper, we propose a comparison of human gesture recognition using data mining classification methods in video streaming. In particular, we are interested in a specific stream of vector of twenty body-joint positions which are representative of the human body captured by Kinect camera. The recognized gesture patterns of the study are stand, sit down, and lie down. Classification methods chosen for comparison study are backpropagation neural network, support vector machine, decision tree, and naive Bayes. Experimental results have shown that the backpropagation neural network method outperforms other classification methods and can achieve recognition with 100% accuracy. Moreover, the average accuracy of all classification methods used in this study is 93.72%, which confirms the high potential of using the Kinect camera in human body recognition applications. Our future work will use the knowledge obtained from these classifiers in time series analysis of gesture sequence for detecting fall motion in a smart home system.

Automatic multiple Kinect cameras setting for simple walking posture analysis

We propose an automatic setting of multiple skeletal tracking Kinect cameras, in lieu of mere using a single camera, to capture a human skeleton because of possible viewing occlusions. Using multiple cameras from different angles gives a more complete whole body; however, more required steps are needed in combining multiple skeletons into one final skeleton. One camera is used as a reference for the other cameras to transform their coordinates into the reference cameras coordinate system. Once every view is in the same coordinate, one skeleton is able to be composed. Due to cameras sensory errors, nevertheless, the supposedly same joint of the skeleton, which is obtained from the transformations, may not be exactly located at the same position. Therefore, average joints are used for the composed skeleton. The skeleton is then used to analyze the walking posture of a human subject in order to check whether or not the walking is balanced.

Survey of Smart Technologies for Fall Motion Detection: Techniques, Algorithms and Tools

The aging population has become a world-wide social concern. The number of people living alone and experiencing falls is increasing. This is a major health risk, especially among the elderly; thus, the early detection of fall motion is of great significance. A smart home care system is needed to monitor abnormal events. This paper first conducts a survey of existing smart systems and techniques in detecting fall motion in human movement, including the emergence of new natural user interface (NUI) devices and systems in the consumer market. Secondly, the paper categorizes smart technologies for fall motion detection into three main technological groups: acoustic and ambient sensor-based, kinematic sensor-based, and lastly the computer vision and NUI. An insightful discussion of each category’s advantages and disadvantages is provided. The findings show a promising research direction of integrating the computer vision with the novel consumer-grade NUI device, such as Kinect, in achieving of an affordable and practical smart home fall motion detection system.

A Framework of Multi-Stage Classifier for Identifying Criminal Law Sentences

Abstract This paper proposes a framework to identify the relevant law articles consisting of sentences and range of punishments, given facts discovered in the criminal case of interest. The model is formulated as a two-stage classifier according to the concept of machine learning. The first stage is to determine a set of case diagnostic issues, using a modular Artificial Neural Network (mANN), and the second stage is to determine the relevant legal elements which lead to legal charges identification, using SVM-equipped C4.5. The integrated multi-stage model aims at achieving high accuracy of classification while reserving “arguability”. Hypothetically, mANN handles well for digesting complexity in case-level issues analysis with acceptable explanatory power and C4.5 addresses the lesser extent of contingency and provides human-interpretable logic concerning the high-level context of legal codes.

Smart bedroom for elderly using kinect

Ideas of building smart homes have been around for many years. However, most of them have been confined to laboratories or specially built, expensive homes due to the high cost of sensory hardware and software. In order to make some of those ideas realized for average-income households, we could resort to lower cost devices and focus more on smaller space like a bedroom or living room. Kinect sensors are able to capture motions of a human and present them as a series of time-dependent skeletons. They have been utilized in many fields such as entertainment, security, and health care. Particularly, they could be applied to assist the elderly population of which the number has been increasing substantially around the world when compared to other age groups. This work proposes the simple yet effective system that would facilitate the daily activities of an elderly person in a bedroom, e.g., helping the elderly turn on or turn off electric devices such as a television, electric fan, or room light without touching them or their remote controls. The system could also detect and alert the elderly of falling out of bed and monitor signs of abnormity or any need of assistance so that family members or caregiver could be timely notified.

A genetic algorithm for the multi-class contingent bidding model

Abstract.This paper proposes a Genetic Algorithm (GA) in searching for a near-optimal sequence of jobs in a make-to-order (MTO) production system in order to maximize the average marginal revenue earned per bid in the bidding model that allows contingent orders. Even though the complexity of the sequencing problem is NP-hard by nature, it is found to be a key determinant in improving the capacity allocation and the expected tardiness cost for an arriving order. The model incorporates operational constraints and marketing policies to effectively reflect the interests of customers. A simulation study was conducted to analyze the relative performance of the proposed system in a finite horizon. The results show the significant impact of the ordering sequence on the average marginal revenue and that the GA is an effective and efficient method to search for a good sequence and can improve the profit margin of the MTO firm and satisfaction of its customers.

Factors Affecting Student's Intention to Choose IT Program

Abstract This paper proposes a two-fold study: (1) to find the factors affecting attitude of Thai students for choosing Information Technology (IT) program and (2) to investigate the existence of gender gap in behavioral intention. The study is based on the Theory of Reasoned Action (TRA) as a theoretical framework. The factors that may affect students’ behavioral intention to choose IT program are categorized into two dimensions: attitudes toward choosing IT program and subjective norm. The web-based questionnaire is employed to collect data from a sample of 67 local Thai Grade 12 students of both genders who have intention to study in IT undergraduate program at School of Information Technology (SIT), King Mongkuts University of Technology Thonburi (KMUTT). The result of statistical analysis shows that TRA is effective for explaining the behavioral intention. Male and female students hold the same set of attitudinal attributes when deciding to enter IT program, hence, an IT school shall implement common strategies to grasp intention from both genders. The most effective strategy to gain students intention is to build up the reputation of IT program.

IDENTIFICATION OF CRIMINAL CASE DIAGNOSTIC ISSUES: A MODULAR ANN APPROACH

A knowledge discovery model has been developed to manage the facts discovered in criminal cases in the court of law and to identify the relevant diagnostic issues. This study focuses on the offence against life and body section of the criminal law codes of Thailand. To identify the criminal case diagnostic issues, a set of artificial neural networks (ANN) classifiers is heuristically configured and modularly organized to operate upon the discovered facts. This modular network of ANNs forms an effective system in terms of determining power and ability to trace or infer the relevant reasoning of such a determination. Experiments have been conducted to demonstrate the applicability of ANN for various case studies and to generate comparative results for providing insights into both technical and legal aspects of these cases. In this study, a modular ANN with the support of Principal Component Analysis (PCA) as an automatic input selection mechanism provided the best results with accuracy up to 99%, using 10-fold cross-validation. A sample case is included to illustrate the effectiveness of the proposed system.

Post-Fall Intelligence Supporting Fall Severity Diagnosis Using Kinect Sensor

This paper proposes a fall severity analytic and post-fall intelligence system with three interdependent modules. Module I is the analysis of fall severity based on factors extracted in the phases of during and after fall which include innovative measures of the sequence of body impact, level of impact, and duration of motionlessness. Module II is a timely autonomic notification to relevant persons with context-dependent fall severity alert via electronic communication channels (e.g., smartphone, tablet, or smart TV set). Lastly, Module III is the diagnostic support for caregivers and doctors to have information for making a well-informed decision of first aid or postcure with the chronologically traceable intelligence of information and knowledge found in Modules I and II. The system shall be beneficial to caregivers or doctors, in giving first aid/diagnosis/treatment to the subject, especially, in cases where the subject has lost consciousness and is unable to respond.

Sentence identification system based on criminal law ontology

ABSTRACT In this research, an identification system is proposed for the benefit of determining a possible sentence and retrieving the related cases under criminal law codes of Thailand. The system is based on the developed criminal law ontology for the advantage of structuralizing and semanticizing in selected articles. From the underlying legal elements described in law codes textually, the ontology shall provide the determination of possible sentences which consist of judgment and theoretically figured range of punishments. An application has been designed and demonstrated in two consequential modules: the legal elements identification and the sentence identification. The developed ontology and its extended web application will be shown and illustrated as a sequential flow and evaluated by the legal experts and the end-users. The evaluation results showed that the averaged satisfactions for both groups of experts and end-users were 89% and 84%, respectively.