這是一個資訊化以及數位化的時代,各式各樣的科技產品皆不斷的翻新以及進步,其中聲音以及音訊的處理一直都是大家所關注的焦點,幾個大方向包含了語音編碼(Speech Coding)、語音辨識(Speech Recognition)、聲學回音消除(Acoustic Echo Cancellation簡稱AEC)以及用來針對雙方通話改變的機制,雙邊通話偵測(Double Talk Detection 簡稱DTD),其中後兩者是本論文所要探討的重點。 早期技術尚未發達,通話之間所造成的回音問題常常造成很大的困擾,使通話的品質大大的降低,於是可適性濾波器加上合適的聲學消除演算法相互搭配進而變得格外重要,但是回音消除的演算法層出不窮,哪一種在使用上最有效益,則是我們所要研究的,本研究透過MATLAB程式模擬探討幾種回音聲學消除演算法以及搭配雙邊通話偵測演算法的效果,並且套用在各種不同的情況做了比較,包含AEC演算法的收斂速度、聽覺品質以及計算複雜度,讓我們了解其在未來的使用上該如何抉擇。
This is an information and digital era. Many technology products are improved. However audio and speech process are important focus. Research directions include speech coding, speech recognition, acoustic echo cancellation (AEC), and double talk detection (DTD) in an AEC system. The latter two of are the focus discussed in this paper. Early technology is not yet matured. The echo in communication may cause severe problem on the perceptual quality. So adaptive filter and suitable AEC algorithm jointly become particularly important. Since there are so many echo cancellation algorithms, which one is the most beneficial is what this paper concerned. Using the MATLAB as simulation program, this research applied three acoustic echo cancellation algorithms and one double talk detection algorithm. We investigate a variety of conditions including the AEC convergence rate, the perceptual quality and computational complexity, so that we can provide useful information in algorithms selection.