應用支撐向量機於多點觸控手勢辨識系統

Applying Support Vector Machine on a Recoginition System of Multi-touch Gestures

戴孝樺(Hsiao-Hua Tai)

指導教授 ： 劉傳銘

國立臺北科技大學/電資學院/電資碩士班/碩士(2010年)

https://doi.org/10.6841/NTUT.2010.00590

摘要

本論文的主要目的在於如何應用SVM(Support Vector Machine)來設計一個兼顧高辨識率及高效率的多點觸控手勢辨識系統。隨著電腦科技的進步，傳統的鍵盤、滑鼠輸入方式已經漸漸不敷使用，而語音、攝影機、觸控面板等等的輔助輸入方式則愈受重視。傳統的觸控板，多是單點輸入，僅能取代傳統的滑鼠，但近年來，多點觸控板技術成熟，如何辨識多點輸入成為最近研究的熱門主題。本論文以SVM(支撐向量機)為基礎，另外針對SVM的特性及需求，加以研究，設計了一個可以辨識多種手勢而且高效率的系統。
根據研究指出，支撐向量機(SVM)是一個相當有效的分類工具，可是SVM在運作的過程中，所需要輸入的資料，必須為固定維度，這在實務中，常是個大問題，因為需要分類的資料，有可能是維度不固定，因此本論文，主要就是研究如何將維度不固定的資料，抓取其特徵值，轉化為固定維度的資料(在此稱為特徵向量)，用以當做SVM的輸入資料。本研究發現單純的取樣方法，辨識的效果及效率都是最佳的。

關鍵字

SVM ； Support Vector Machine ； 支撐向量機 ； 多點觸控

並列摘要

The main purpose of this paper is to learn how to use (apply) SVM (Support Vector Machine) to design a High Recognition Rate and high efficiency multi-touch recognition system. With the advances of computer technology, the traditional data input methods, such as keyboard and mouse, have become insufficient, while the voice, camera and touch panel, the new data input methods are getting more attention nowadays. A traditional touch pad, with single-point input can only replace the traditional mouse. In these days, multi-touch panel technology is getting mature and reliable, therefore, how to identify multi-point input has become a hot research topic recently. This paper is based on SVM (support vector machine), in addition to discuss and study more additional features and needs of the SVM, to design a algorithm can recognize variety of hand gestures and highly efficient.

According to the study, the support vector machine (SVM) is a very effective tool for classification, however, during the process of SVM, the enter data has to be a fixed dimension, which will be a big problem in practice, since some data dimensions are not fixed. Therefore this paper is to study how to sort out and use the feature value when dimension of the data is not fixed, then transfer it to a fixed dimension data (here referred to as feature vectors), so it can be used as the input data for SVM. The study found that a simple sampling method, identify the effectiveness and efficiency are the best ones.

並列關鍵字

SVM ； Support Vector Machine ； Multi-touch

參考文獻

[11] Hsiang-Chuan Liu,Kuei-Kuang Lin,You-Ren Lin and You-Ren Lin.（2007）＂Choquet Integral Regression Models Based on Polyvalent Fuzzy Measure＂, Available at Conference On Business Operations and Management.

[1] Cortes, C. and V. Vapnik.（1995）"Support-vector networks", MachineLearning 20, 273–297.

[2] 邱雅卿(2008) ＂無參數加權特徵萃取支撐向量機運用於甲狀腺分類＂，亞洲大學生物資訊學系碩士論文。

[3] 陳國鴻（2005）＂導入權重概念對支撐向量機分類正確率之影響＂，國立成功大學工業與資訊管理學系碩士論文。

[4] Chih-Wei Hsu, Chih-Chung Chang and Chih-Jen Lin.（2008）"A practicalguide to support vector classification" Available at http://www.csie.ntu.edu.tw/~cjlin/papers/guide/guide.pdf.