利用隱藏式馬可夫模型辨識獨立字之研究

Abstract

在資訊科學領域裡，語音辨識佔有重要的地位，此因語者在任何地方不需 用眼睛或手即可與電腦溝通，甚至也不必透過鍵盤，滑鼠等週邊設備。近 年來，隱藏式馬可夫模型（ＨＭＭ）已被廣泛地應用在資訊科學領域，特 別是語音辨識方面。在本論文中，我們對隱藏式馬可夫模型的基本理論與 演算法做深入的介紹，並且利用此模型發展出獨立字之辨識系統，包括（ １）中文字之辨識系統及（２）語者辨識系統。在簡單的環境下，如小字 彙或少數語者的情形下，已有不錯的結果。我們將就這些研究結果作簡單 地討論，以探討未來解決一般語音辨識問題之方法。 Speech recognition plays an important role in computer science because the voice input to computers has the advantages of of offers fast, hands free, eyes free and location free. The communication with computers can be carried out convenient without the use keyboards or mouse. Hidden Markov models(HMM) has been using widely in many computers science areas, especially in the area of speech recognition. In this theis, the basic theories and algorithms about HMM are described in detail. Consequently, an isolated-word speech recognition system based on HMM is implemented in this thesis research. Applications of the implemented system in (1)isolated-word recognition for Mandarin, and (2)speaker identification is investigated. Satisfactory results have been obtained for some simple cases, e.g. small vocabulary, small number of speakers, etc. Possible extension of the research results to solve more general speech recognition problems is briefly discussed.