頭部姿勢辨識應用於游標與機器人之控制; The Cursor and Robot Application via Head Posture Recognition Control

NCUIR > College of Electrical Engineering & Computer Science > Graduate Institute of Electrical Engineering > Electronic Thesis & Dissertation > Item 987654321/10049

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/10049

Title:	頭部姿勢辨識應用於游標與機器人之控制;The Cursor and Robot Application via Head Posture Recognition Control
Authors:	侯俊成;June-Cheng Ho
Contributors:	電機工程研究所
Keywords:	足型機器人;線性區別分析;主成份分析;人臉偵測;影像處理;殘障輔具;Image Processing;Face Detection;Principle Component Analysis;Assistive Technology;Hexapod Robot;Linear Discriminant Analysis
Date:	2006-06-20
Issue Date:	2009-09-22 12:04:48 (UTC+8)
Publisher:	國立中央大學圖書館
Abstract:	本論文提出一個利用頭部轉動姿勢控制滑鼠游標與足型機器人的方法，主要是藉由影像處理技巧與多變量分析法中的費雪線性區別分析法(Fisher Linear Discriminant, FLD)來抽取出頭部姿勢特徵並辨識頭部姿勢，利用所辨識出的頭部姿勢結果，如頭部正面、朝上、朝下、朝左、朝右，分別控制電腦滑鼠游標的停止、往上、往下、往左、往右移動或機器人的停止、前進、後退、左轉及右轉的動作。並且希望藉由本系統能夠幫助一些肢體行動不便的人士，尤其是手部或足部有功能障礙者，使其能夠輕鬆地操作電腦，更便利於與外界溝通，改善生活品質。在本論文中，使用對光線強弱較不敏感的YCbCr(Luminance, Blueness, Redness)色彩空間來偵測膚色區域，且由於一般使用者使用電腦時，大多與電腦距離很近，所以人臉膚色區塊會在輸入影像中佔最大區域，因此我們利用區塊標記法(connect component labeling)，快速地定位出最大膚色區塊，則可定位出實際的人臉影像，然後利用此人臉影像與線性區別分析法所訓練出來的訓練特徵參數，以歐氏距離(Euclidean distance)為決策法則進行比對，並輸出結果，以控制滑鼠游標與機器人移動。 The goal of this paper is to develop a head-pose recognition system to control the computer cursor and a hexapod robot. The techniques of image processing and Fisher linear discriminant analysis are used to recognize the pose of head, such as front, up, down, left and right. When the pose of head is recognized, this result is applied to control the cursor or the hexapod robot. This system can help some disabled people who are impaired in hands or feet to operate the computer and to communicate with the external world. This technique would improve the quality of their life. In this paper, the YCbCr color system is adopted to detect the color of skin, because YCbCr is not sensitive to the variation of light source. When using computer, most of users are close to the camera which is equipped in front of the computer. Thus, the face region would occupy the most area in the captured image. We use the method of the connect component labeling, which can detect the largest region of skin in the captured image and locate the region of human face. Then, the linear discriminant algorithm is applied to extract the feature parameter from the training images and the test image. Finally, by comparing the feature extracted from the captured image to the parameter extracted from the training image, we can estimate the direction of head and control the cursor or a hexapod robot.
Appears in Collections:	[Graduate Institute of Electrical Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Size	Format

社群 sharing

Loading...