- 學位論文
雲端字型服務系統之字型佈局調整與字型風格分類
Adjusting Font Layout and Classifying Font Style for Cloud-Based Font Service System
摘要
網路科技發達，網際網路成為最常被使用的溝通與表達管道之一，許多客製化需求日漸擴大，相關服務也因應而生。文字為古往今來重要的溝通媒介之一，也成為人們展現自我內涵與思考的媒介，字型檔的個人化與客製化已成為一種趨勢。現今提供個人手寫字型服務的JustFont字型公司，需花費較高額的金錢與等待時間，服務門檻較高。王之盈在「中文字型產生與部件組字之分析(2013)」中提出的雲端字型擴充服務架構，主要目的即是以低門檻簡便的方式與流程讓使用者可自行創建以漢字為主的個人化手寫字型，期許將創字、管理、應用三部分串連起來成為網站的服務流程。此服務目前尚有兩項主要缺陷，一是透過手寫輸入的字形雖然具有個人特色，但也容易產生位置飄移不定或粗細不一等現象。二為當使用者在平台上建立的字形資料量越來越大時，這些字體的管理與分類需要耗費使用者相當大量的時間與心力。 本論文提出字型佈局調整與字型風格分類來解決上述之問題，字型布局調整主要目的為透過自動化字型布局調整取代人工調整，調整規格包含字體位置、大小及其粗細，藉此達到解決或減輕手寫字問題;字型風格分類則是希望以篩選過的特徵值並投入SVM中，訓練出字型風格分類器，自動推薦使用者適當的分類，減輕使用者管理字型之負擔。在實驗結果方面，字型布局調整之部分，以問卷做為檢測使用者認同度之管道，統計問卷後得到82%之使用者認同本實驗效果，驗證了本研究的可用性。而在字型風格分類上，在10組字型遞增測試下，最高辨識率可達65%。期許以本研究提出之解決方式更完善雲端字型服務之功能，貼近使用者需求並促進個性化字型的發展。
並列摘要
Due to the development of technology, more and more services offering personalized products have been emerging, such as customized personal fonts. Handwriting is a miniature of personality and often reflects one’s style. JustFont provides a service for users to customize their own handwriting Chinese fonts. However, the service costs a lot of money and time. Chih-Yin Wang had proposed a Cloud-Based Chinese Font Service to help users create and use customized Chinese font simply and easily. The service is consisted of font creation, management, and application. But there were still some unsolved problems. For example, the handwritten characters are probably too larger or smaller in size, improperly aligned with others, and with strokes of different thickness. Furthermore, the management model is inconvenient in long term for users when the number of handwritten characters is large. In this research, we propose several methods to automatically adjust the layout of the handwriting fonts. A questionnaire was made to assess the adjustment. On average, 82% of respondents agreed that the sizes of the adjusted characters are more appropriate and the thickness of strokes is more consistent. As well, the adjusted characters are better aligned with each other than the original characters. To manage characters, users can assign styles to every character. A classification subsystem was built to help users assign styles to characters. Based on the features extracted from the handwritten characters in the database, a support vector machine (SVM) was trained to assign the character style to a new arrival. In the experiment, when Chinese characters from ten handwriting fonts were used, the classification accuracy of the SVM was 65%. The experiment result shows that the proposed classification subsystem can help users manage their own handwritten characters effectively.