Equtatetor-新一代智慧型數學處理器
此研究的目的是要設計出一套完整編輯顯現數學式、加以計算，並求出解的一套方法與成品。而這項工作的執行者，在此稱之Equatetor 。一般的數學式子，若要計算的話，普通的計算機是不足夠的。原因是它們沒有辦法表現出數學式的「原貌」，例如分號、指數、函數、根號等數學符號混在一起時的情況。於是，我便擬定了一個研究，希望設計出一套更方便且實用的方法。換句話說，我要設計出一個功能強大的工程計算機程式。其中，自然牽扯到數學式子的顯現方式（以MathML 實現），以及計算機科學的演算法及資料結構。我主要的目的有四：(1) 顯示數學式(2) 方便編輯數學式(3) 計算數學式(4) 處理可以以不同形式輸出解答的計算（如輸出分數、根號、函數解等）。研究結果中，成功地運用XML 中的MathML 與二分逼近分數等演算法及若干資料結構，達到了以下實用的幾點：(1) 結構化的數學式編輯(2) 完整地顯示數學式(3) 正確運算並輸出運算式的答案(4) 提供一般數學形式之解（非小數之解）;The object of this study is to design a method and processor which is able to edit, display a mathematical expression representing a number, calculate and output the answer. The executor of this task is called Equatetor. Normal calculators are not adequate for this kind of task. The main reason is that they can’t reveal the original expression, such as fractions, radicals, exponents or mathematic functions. Therefore, a simple and convenient method is needed. To perform the possible way of handling those tasks, a computer program has been written. Several techniques were used, such as MathML, computing algorithms, data structures, and so on. Following are main purposes: (1) Displaying mathematical expressions. (2) Editing mathematical expressions simply. (3) Calculating mathematical expressions. (4) Outputting the answers(in different expressions). And the achievements:(1) Structured methods of editing of mathematical expressions. (2) Displaying mathematical expressions completely. (3) Calculating mathematical expressions precisely. (4) Offering answers in different expressions.
排印OnLine-PHP 在文字圖形的應用
從小習染之餘，對篆刻藝術的濃厚興趣及對電腦程式之熱愛，萌生將篆刻藝術與電腦 科技結合，於是應用PHP（Professional Home Page）的文字圖形（Text Image）處理功能 將中華文化之篆刻藝術透過網際網路推廣到世界各角落。本研究規劃建置了一個 FreeBSD Server 可完善處理PHP 功能的作業平台、應用PHP 的文字圖形處理功能，編 寫可在線上DIY 設計印章的程式，以及藉由Java Script 的技術在Client 端建立一個人 性化且相容性高之使用者界面網頁，達到提供多語系、多功能、依使用者偏好作調整 印章及輸出分享的功能。為使非漢語系國家能深刻體驗中華篆刻藝術，本研究同時建 立一個超過三萬筆之由英文名音譯中文名的資料庫，提供給使用者切身的服務。 Inspired by the interest of the art of seal engraving and the love for the computer programming beginning from my childhood, I combine the art of seal engraving and computer technology, and apply the Text-Image's function of PHP to promote the seal-engraving art of Chinese culture via the Internet to everywhere around the world. This project constructs a FreeBSD Server to process PHP platform perfectly, applies the Text-Image's function of PHP to write a program for online DIY pattern design, and uses Java Script to establish a human and compatible user-interface web page for clients to provide multi-language, multi-function, and being able to adjust pattern according to the user's preference and output sharing functions. This project also sets up a database of English-Chinese translation of over thirty thousand English terms to help people who are not Chinese to experience the art of seal engraving.
Digital Viedo Compression Enhancement With Reduced Psychovisual Redundancy
Video compression is indispensable to web streaming and memory storage.Most video\r compression technology has difficulty to achieve high quality video at lower bit\r rates.Apparently,limited transmission bandwidth and network resources often degrade\r video signals.Thus the goal of my research was to enhance video degrade video\r signals.Thus the goal of my research was to enhance video compression performance and\r to improve visual quality.It is hypothesized that the reduction in neighboring pixels\r coding,and humans perceptual mechanisms(psychovisual)redundancy could produce a\r low-complexity geometry streams for animated visual objects.A set of algorithms is\r developed to parse bidirectional interpolation pixels into their characteristic cells,which\r vary in spectral energy and wavelength.The bits contained in these cells are vectorized and\r transformed recursively to identify lower correlations among vector arrarys for blocks\r filtering.DCT function calculates energy ratios between high spatial frequency and low\r spatial frequency,to devote most of the highest spatial frequency bits with the calculated\r energy ratios.A variable quantization method is used to measure the sensitivity of colors\r and its intensity ratios to restore any missing high spatial frequency pixels.presnted in\r mathematical intrinsic.This approach leads to the ability to compress video data that\r normally require a large amount of memory to store and high bandwidth to\r transmit,Results form the enhanced video compression experiment have attained\r 0.1bpp(256kbps,25fps)without noticeable effects comparable to the video compression\r technique that achieved 0.5bpp(1.5Mbps,25fps)in use today.
Computer Vision for Alternative Input Systems
In the fast-paced environment of a hospital intensive care unit (ICU), good doctor-patient communication is essential. However, medical conditions and devices often inhibit a patient’s ability to speak, write or type. Current assistive communication devices are either prohibitively expensive or cumbersome and time-consuming, creating a gap in communication during a patient’s first days in the ICU. This project applies computer vision to develop a low-cost software solution that bridges this gap by enabling patients to generate words with eye movements. In the system, a webcam acquires an image of the patient, and an image processing algorithm classifies patient’s gaze as pointing in one of eight directions. Each direction corresponds to an option on a graphical menu presented to the patient on the computer’s display. The patient can use the menu to select a preformed phrase from a list of common phrases. Patients desiring to express more complex ideas can type custom words using the menu as an ambiguous keyboard (similar to a phone keypad). In either case, the patient-generated text will be displayed on screen and read aloud through the computer’s audio system. The only hardware requirements are an existing computer and a $6 webcam. The program can process and respond to an image in 148ms. A new user can be trained in approximately 10 minutes, and after training can type a simple phrase such as “hello world” in 40 seconds. While further testing and improvement is required before the system will be ready for implementation, the project shows promise as a low-cost solution to ICU communication.
利用Google Maps 建立高中地理資訊系統
台灣通史「婆娑之洋，美麗之島」，身為台灣人，應知台灣事，本研究本維 基百科（Wikipedia）所揭示的「海納百川，有容乃大」精神，以Apache 網頁伺 服器為平台，MySQL 為資料庫，利用AJAX、PHP 網頁技術，並使用Google Maps 提供之服務，建立符合Web 2.0 精神的的b04地理資訊共享系統。在此系統下， 我們允許教師(編輯者)以合作方式上網提供與課程相關的地理資訊，並可進行資 訊修改或刪除，希望集合所有教師之力，提供給學生最豐富、最正確的b04地理 資訊系統。 In this paper, we propose a new map-based collaboration environment “MyGIS” for geographical information distribution. In MyGIS, the publishers/teachers can easily collaborate to annotate a spot with contents such as messages and links. The annotations of a spot are updated on the map in real-time which enables the receivers/students to correctly study at anywhere and anytime. To ensure the trustfulness of the content, the publishers are authenticated using login mechanism. In this study, several technologies inc lu ding Apache, PHP, MySQL, Ajax are involved. Finally, We implemented MyGIS on Google Maps and evaluated its feasibility and usefulness.
應用生物特徵及視覺密碼學防治車輛失竊和酒醉駕車
車輛遭竊以及酒醉駕車對社會安全造成的危害普遍存在於世界各先進國家。以目前看來，此問題並非嚴刑峻罰就能解決的。若本研究計畫將來普及應用於大眾，必定可以同時減少車輛遭竊之機會和預先防範酒醉者開車駕駛。 生物特徵泛指指紋、虹膜、臉型、聲紋甚至DNA 等等，它的特性即每個人均具有此獨一無二的性徵，加以應用就如同無法被仿造或偷竊的身分證明。若車主能利用自己的生物特徵為車輛上鎖，即可防止車輛遭竊。 視覺密碼學是屬於資訊隱藏技術的一種，藉由兩張以上的影像疊合後呈現其中隱藏之機密資料。雖然視覺密碼學原先純粹是為了加密資料，但我們將之加以修改，設計出複雜且無意義的偽裝影像，使其不易被酒醉後神智不清的人疊合成功而讀出其中的加密資料。 結合生物特徵及視覺密碼學，即可設計出一套可安裝於車上的系統，不僅防止車輛失竊又能防止酒醉駕車事故的發生。 ;The damage which is caused to social security by car stealing and drunk driving universally exist in others developed countries. As things are, these problems can not be solved by severe law. If our project can be applied in the masses extensively, it must be able to prevent both car stealing and drunk driving. Roughly speaking, biological characteristics mean fingerprint, iris, face, pronunciation and DNA. Their uniqueness is a special feature that everyone has. If we can apply it more, the characteristics are identification that can not be imitated or stolen. If the car owner can use his or her biological characteristics to lock the car, the car stealing could be prevented. Visual cryptography belongs to a kind of information hiding. By stacking two or more images, user can recognize the secret information hiding in the stacking. Although visual cryptography is merely created for encryption at first, we modify it and design a complex as well as meaningless stego-image. That makes the unconscious driver who is drunk fail to both stack the images and recognize the secret information hiding inside. We combine biological characteristics with visual cryptography and design a suit of system that is able to install in the car. This system can prevent not only car stealing but also drunk driving.
灰階影像內容之檢索技術
資訊時代的來臨，促使我們的社會型態大幅改變，無一不朝數位化的方向邁進。網際網路與資訊科技的快速發展，近年來影像資料庫和數位圖書館大量的成立，關於影像資料檢索的研究已漸漸成為一門極為重要的研究議題。在本研究報告中，我們提出一種植基於向量量化編碼法的灰階影像檢索技術。向量量化編碼法是一種極為簡單的影像壓縮技術。我們應用這個編碼法，來萃取出灰階影像內容的特徵值。同時我們也計算出整張影像與影像中間位置的像素平均值，作為日後檢索影像時過濾掉影像資料庫中不需要比對特徵值的依據。\r 我們所提出的方法能有效地萃取影像內容的特徵值並且讓使用者可以快速且正確地查詢到所需要的影像。當影像資料庫中存在與查詢影像完全相同的影像時，我們的檢索技術都能在第一時間第一順位檢索出這張影像。即便影像資料庫中不存在與查詢影像完全相同的影像時，我們的檢索技術平均74.3%也能在第一時間前五個順位檢索出最相似的影像。\r \r With the coming of the information age, our sociological system change extensively, and everything has moves toward digitization. Due to the rapid development of Internet, information technology, the rapid growth of image databases and digital libraries recently, the related researches of image retrieval have become a very important issue. In this memoir, we propose an image retrieval scheme based on the vector quantization to retrieve similar images from the image database according to the pre-collected image features. Vector quantization is a very simple image compression scheme. We have applied vector quantization to extract features from grayscale image. In order to speed up the retrieval process, we also calculated the mean pixel values of whole images and the central part of the image to filter the images, which are significantly dissimilar to the query image.\r The experimental results show that our proposed approaches can effectively extract features from the image and enable users to retrieve images from image database quickly and accurately. When images stored in image database match query image, the proposed scheme can instantly retrieve the stored image at the first rank. Even though images stored in image database query image exactly, the proposed scheme can instantly retrieve the stored image over 74.3% at the first five ranks.