28.11.2014 Views

Preface - kmutt

Preface - kmutt

Preface - kmutt

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

80<br />

MOVIE CLIPS USING MOVEMENT AND<br />

PACE FEATURES<br />

Saowaluk Watanapa, Bundit Thipakorn,<br />

Nipon Charoenkitkarn<br />

The 5 th IASTED International Conference on<br />

Visualization, Imaging, and Image Processing<br />

(VIIP2005), September 7-9, 2005, Benidorm,<br />

Spain<br />

Classifying multimedia data into<br />

semantic, such as emotional, categories is<br />

challenging and useful. Using two film artistic<br />

components: movement and pace, this paper<br />

proposes a method that can distinguish movie<br />

clips belonging to the excitement class from the<br />

others. Movement is represented by the Average<br />

Squared Motion Vector Magnitude feature, and<br />

pace is represented by the Average Shot<br />

Duration feature. Classification experiments<br />

with 101 data clips, excerpted from 24<br />

Hollywood movies, are conducted, employing<br />

minimum distance and k-NN classifiers. The<br />

results show that the selected features can<br />

potentially separate the excitement movie clips<br />

from the others with above 90% accuracy.<br />

IC-005 SEMANTIC PERSONAL IMAGE<br />

CLASSIFICATION BY ENERGY<br />

EXPENDITURE<br />

Sirinporn Chinpanchana,<br />

Songrit Maneewongvatana, Bundit Thipakorn<br />

International Symposium on Communications<br />

and Information Technologies (ISCIT2005),<br />

October 12-15, 2005, Fragrant Hill Hotel,<br />

Beijing, China, p. 1072<br />

Semantic personal image classification<br />

is an attention problem in multimedia image<br />

retrieval. In our previous work, we classified<br />

semantic images into Business, Leisure, and<br />

Sport categories by integrating the frequency<br />

pattern relationships between body parts and<br />

objects. However, the accuracy mainly depends<br />

on their objects. In the images that have high<br />

semantic complexities, the body movement play<br />

important solve on the meaning of image. In this<br />

paper, we present a new model to achieve<br />

more effective classifier called an Energy<br />

Expenditure model (EE). The EE model is based<br />

on the concept that human subjects in different<br />

classes of images are likely to spend different<br />

amounts of energy. The angular position and<br />

flexion forces are related into each body part.<br />

KMUTT Annual Research Abstracts 2005<br />

Experimental results show that the EE a can<br />

achieve an improvement of semantic images.<br />

IC-006 THE DEVIATION OF GLOTTAL FLOW<br />

DERIVATIVE MODEL DUE TO EMOTIONAL<br />

SPEECH<br />

Suthathip Chuenwattanapranithi,<br />

Songrit Maneewongvatana, Bundit Thipakorn<br />

The 2005 Electrical Engineering/Electronics,<br />

Computer, Telecommunications, and<br />

Information Technology International<br />

Conference (ECTI-CON 2005), May 12-13,<br />

2005, Asia Pattaya Beach Hotel, Pattaya,<br />

Cholburi, Thailand<br />

It is well studied that the changes of the<br />

glottal flow pulse have a direct effect on the<br />

acoustic and perceptual features of an<br />

individual's utterance. These features, in turn,<br />

have influences on perceptual voice qualities of<br />

the phonations which are closely related to the<br />

emotion of the speaker. In this paper, we study<br />

the effects of emotional speech on the shape of<br />

the glottal flow pulse. Specifically, we use LF<br />

model to represent the glottal flow derivative<br />

pulse of emotional speeches. Our hypothesis is<br />

that each emotional speech would have a distinct<br />

shape of the glottal flow derivative pulse<br />

compared to neutral speech. This implies that<br />

parameters of the LF model can be used to<br />

determine the emotion of the speaker. We<br />

perform a set of experiments on several<br />

emotional speech databases, each provides four<br />

emotional states: anger joy, sadness and neutral.<br />

Two-way ANOVA analyses confirm our<br />

hypothesis. The essential LF parameters are<br />

extracted and used for emotional speech<br />

recognition. This feature set achieves high<br />

accuracy on recognizing anger and sadness<br />

emotional speeches.<br />

IC-007 A COMPARISON OF TWO CAMERA<br />

POSE METHODS FOR AUGMENTED<br />

REALITY<br />

Varin Chouvatut, Suthep Madarasmi<br />

The 7 th IASTED International Conference Signal<br />

and Image Processing, August 15-17, 2005,<br />

Honolulu, Hawaii, USA., p. 554<br />

In an augmented reality (AR) system,<br />

also known as mixed reality (MR) system, 3-D<br />

virtual objects can be realistically added into a<br />

real, dynamic, 3-D environment to create an<br />

International Conference

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!