本論文的研究主題是針對前人的關鍵詞萃取中的特徵參數擷取作改良，將前人所用之LPC方法改為MFCC方法，並結合語音辨識系統建構一套聲控車系統。本論文主體可分為兩個部分，在關鍵詞萃取部分，關鍵詞與無關詞模組是用次音節模型來建立，目的是使的系統更有可攜性。第二部分是將建立出來的模型，利用Visual Basic 6的開發環境，應用一階動態辨識演算法，將我們的辨識技術製作成視窗化的人機介面，達到即時辨識的效果，並且可以根據辨識的結果，與市售的遙控車結合，讓車子可以依照使用者所講的方向移動。The topic of the thesis is modifying part of keyword spotting that feature extracting, we substitute method Mel-frequency cepstral coefficients for method Linear prediction coefficients, and construct a voice-activated car by speech recognition.There are two topics in the thesis. In the first part, we focus on keyword spotting, and our keyword models and garbage models are building by sub-syllable models, and the advantage is that the system can save a lot of time. In the second part, we use Visual Basic 6 to make a human-machine interface for real-time recognition, and we combine the human-machine interface with remote control car to make a voice-activated car.
