Detail Hand gesture recognition and face detection in images Bibliografi Author: Yang, Ming-Hsuan ; Ahuja, Narendra (Advisor) Topik: COMPUTER SCIENCE Bahasa: (EN )    ISBN: 0-599-76367-1     Penerbit: UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGN     Tahun Terbit: 2000     Jenis: Theses - Dissertation Fulltext: 9971227.pdf (0.0B; 5 download) Abstract The goal of this thesis is to develop vision-based interfaces between man and machine. Various aspects of research on intelligent human computer interaction are addressed in the context of computer vision and machine learning. The first part of this thesis aims to extract two-dimensional motion across image frames and classify underlying three-dimensional motion patterns. We have developed a method for extracting and classifying two-dimensional motion in an image sequence based on motion trajectories. For concreteness, we focus on image sequences showing hand motions for signs in American Sign Language. The same method can be adapted to recognize other motion patterns such as human walking and running. In this context, it is important to identify certain body parts of the signer in the image sequence. Also, it is important to locate the hand position relative to head and the rest of the body. Therefore, a related problem is to detect human faces robustly. In the second part of this thesis, three methods have been developed to detect human faces in color or gray-level images. In the third part of this thesis, we have developed a distribution free learning theory to a visual recognition problem: object recognition. The problem is viewed as that of learning a representation of an object that, given a new image, is used to recognize the target object in it. The learning account is developed within the PAC (Probably Approximately Correct) model of learnability. For this framework to contribute to a practical solution, there needs to be a computational approach that is able to learn the concepts of objects. The evaluation we provide for this framework relies on the SNoW learning architecture that is used in a large scale experiment. Recently, Support Vector Machines (SVMs) have shown great potential in visual learning and pattern recognition problems. However, training a SVM for a large-scale problem is challenging since it is computationally intensive and the memory requirement grows with square of the number of training vectors. In the fourth part of this thesis, we have developed a geometric approach to train SVMs and compared its performance against conventional methods. Opini Anda Klik untuk menuliskan opini Anda tentang koleksi ini! Lihat Sejarah Pengadaan  Konversi Metadata   Kembali

Copyright © 2006, 2007 Unika Atma Jaya, all rights reserved