- 學位論文
以一個隱藏式馬可夫模型與模糊理論用於手寫筆跡心理學的文件處理與分析
Graphological Document Processing and Analysis Using a Hidden Markov Model and Fuzzy Theory
摘要
每個人在成長的過程當中，都會擁有其獨一無二的筆跡風格，在每個就學階段中，所學習到的書寫方式，也都可以反應出當時的心理特徵。手寫筆跡是一種穩定且沉默的特徵，可以反應出書寫者的個人風格。本論文研究如何利用這個特質，發展出從筆跡推斷出心理特徵的自動方法，以便在缺乏筆跡心理學家的情況下，能處理大量受測者的筆跡。 在本文當中，我們就是要利用訊號處理的方法，包括隱藏式馬可夫模型、維特筆演算法與模糊理論，設計出一套辨識手寫筆跡心理學的介面系統，依心理學家的觀點來抽取人類手寫字的特徵，進而由這些特徵來了解一個人的心理和其特性。本研究方法分為三部份，第一部份：使用影像處理與圖形辨識技術，存取文件、文件前處理、文字切割、抽取特徵後分析；第二部份：利用模糊理論的方法，將部份抽取後的特徵做綜合評估；第三部份：配合心理學家專業知識意見與問卷進一步強化及驗證此系統。 實驗結果顯示，本系統可以準確抽取出人類的心理特徵，且平均的心理辨識正確率達到八成零六左右，敏感度為七成二八，有效性為八成四六，Kappa值平均為0.796 > 0.75，系統之可信度非常好。此系統可以有許多應用，例如：在國外已經有很多大企業運用類似的系統，來了解其所需聘用之員工的人格特性，並成為他們徵才選人的一項重要且客觀的參考依據。另外，此系統也可使心理治療師在對病患診治時，當作另一項評估的指標。
並列摘要
Every person is a unique individual who developes his or her own style of handwriting in the process of personal development. The learning of school handwriting style is only a start which may be completely different from his or her later personal style of penmanship. Handwritten style is just like a silent reporter that can reflect the person’s personality. In this thesis, we study how to use this fact to develope a method and tool that can derive a person’s personality automatically based on his or her handwriting style. With this tool, a large quantity of testees can be handled without the presence of graphology experts. In order to facilitate the graphological analysis, we design a human-computer interacted handwriting document analysis system. The system involves the use of signal processing techniques, including Hidden Markov Model (HMM), viterbi algorithm, and fuzzy theory. The system can be divided into three sub-systems, i.e. digital image processing, fuzzy system, and psychological questionary. In digital image processing, we use image processing and pattern recognition method to implement image preprocessing, handwritten character segmentation, feature extraction and feature analysis. For the fuzzy system, we make use of extracted features to perform psychological evaluation. Finally, we use psychological questionary (16 PF), which is based on expertise and knowledge form psychologists, to test and verify the effectiveness of the proposed system. The experimental results show that the successful rate of matching results obtained form psychological questionary is 80.6%. The sensitivity is 72.8% and the specificity is 84.6%. The Kappa rate is 0.796 > 0.75, showing a reliable system design. The psychological human-computer interaction system can be applied to a recruiting process for a company, to a patient evaluation process for a psychiatrist, etc.