[[alternative]]A Study of English Pronunciation Evaluation by an Improved DTW Algorithm

[[abstract]]隨著時代的改變，在全球195個國家中各國家都有著各種語言，各式各樣的口音與學習環境，而大多數的國家之間溝通，幾乎都是藉著en為溝通的橋樑，間接說明著en在正式場合中的重要性與學習需求，本文中經由MATLAB建立系統配合語音評測的方式讓人與電腦溝通作為練習方式，建立一套en練習發音評測系統，取出念出的en轉換成數值，以比對相似程度作為評分系統。 本論文提出9筆定位法可改進動態時間校正演算法(Dynamic time warping ; DTW)中的尋找路徑方式，減少原方法運算次數，減少計算次數量後也相對提高運算速度與標準範例做評分，提供給學習者當作練習en發音的方法與參考依據，經147組評測實驗採用9筆定位的總計算量是原DTW的11~23%，其評測分數在67~92%之間。[[abstract]]As the generation moves on, each country has its own language, accent, and study environment. English is used as the bridge of communication between countries, which shows the importance and the need of English in formal occasions. Use of the communication between human and computer as a way of practice, this paper introduces a system for practicing English pronunciation by combining MATLAB, a system builder, with voice evaluation. This system transfer the English that was pronounced into data in order to create an evaluation base on the similarity of the pronunciation. This paper propesed a modified DTW algorithm to improve the way of finding routes from dynamic time warping (DTW). By decreasing the number of calculation, the calculation speed and standard for evaluation will increase, which can be provided for the participants as practicing the pronunciation and references. Out of the one hundred and forty-seven experiments, the total amount of calculation by using the modified DTW method was 11~23% of the original DTW, Evaluation scores between 67 to 92%