A Web-Based System of Learning Chinese Syllable Pronounce

[[abstract]]為了使能夠在網際網路上達到自我快速學習發音的效果，吾人在網際網路上發展一能夠由老師提供發音，學生上網自行選擇發音和對象來學習，因此於本篇論文中我們將展示一Web-Based華語拼音學習系統。老師先選擇拼音和聲調，再透過ActiveX-Control錄音元件錄製標準音，抽取出特徵參數後傳送至伺服端，並由系統做標準音的模型訓練。而學生選擇拼音和聲調，再選擇有錄製此音的老師，透過錄音元件錄製發音，抽取出特徵參數後傳送至伺服端，由系統根據老師和學生的拼音、聲調資料作計算，評打拼音和聲調分數。在辨識方面，吾人使用隱藏式馬可夫模型來建立辨識模型，根據此模型執行隱藏式馬可夫模型相關的演算法，最後可獲得一個最佳的拼音和聲調模型，吾人使用它來當作拼音和聲調辨識的基準。在評分方面，吾人學生的拼音和聲調最大可能性值，與該老師的標準拼音、聲調門檻值做誤差率計算，以獲得該學生發音的分數。[[abstract]]In order to achieve self studying Chinese pronounce in the internet, we design a system about learning pronounce on the world-wide-web. In addition to run a wave recorder on the web browser, we have to implement an ActiveX record control which records user voice and provide an interface to communicate with the web browser. The system provide user to choice pronounces and tones, then use the ActiveX control to record syllable pronounces, immediately send the parameters of characteristic to the web server. System use the calculated max possibility of pronounces and tones to compare with standard thresholds, lastly present the score. We use Hidden Markov Model algorithm to train the pronounce and tone models, then get the optimal speech characteristic models and save them to the standard pronounce database