Please use this identifier to cite or link to this item:
DC FieldValueLanguage
dc.description.abstract動態時間校準(DTW)演算法廣泛地被應用在語音辨識上。但在用於我們的驗證系統上有幾項缺點:樣本比對時間耗時,臨界點難以設定,系統精確度低。而另一種常用的演算為隱藏式馬可夫模式(HMM),它提供一個相當可靠的方式,並廣泛的應用與整合於語音辨識系統中,但是相對的演算法相當複雜,所需時間與計算量相當龐大,並不適合應用於本系統中,在本論文中,我們結合了動態時間校準演算法,隱藏式馬可夫模式和高斯分佈機率並加上一些前端處理提高辨識率,而使驗證系統能夠在比對時間與系統精確度上取得平衡。 在語者驗證方面,我們拿一未知語者和系統資料庫來做比對。而系統資料庫是藉由一群已知的語者語音樣本架構而成,其特徵向量則是藉由線性預測參數以及梅爾倒頻參數所得。我們將比較兩種不同特徵參數以及狀態數不同時對效能之影響,最後我們將可發現在怎樣的模型下可得一最佳效能。zh_TW
dc.description.abstractDynamic time warping (DTW) algorithm was widely used in speech recognition but it take large computation time and difficult to determine the thresholding value. Hidden Markov Models (HMM) provides a natural and highly reliable way of recognizing speech for a wide range of applications but it is too complex and too time consuming. In the thesis, we take some characteristics from DTW and HMM, and using the Gaussian distribution as front-end processes which can provide a good performance for voice verification. In speaker verification, we take a voice from an unknown speaker to match a set of known speakers from database. Feature vectors are extracted from the voice samples by using Linear predictive coding (LPC) algorithm or Mel-Frequency Cepstral Coefficients (MFCC) algorithm. We compare the outcome of different feature vectors in both LPC and MFCC algorithm and to observe the influence of the state size. The simulation results show that using the LPC algorithm is better than using the MFCC algorithm in terms of the correctness to identify the right person with right password.en_US
dc.description.tableofcontents第一章 緒論 ---------------------------------------1 1.1研究動機 -------------------------------------1 1.2語者辨認概論 ---------------------------------4 1.3章節大要 -------------------------------------4 第二章 語音辨識的理論基礎 -------------------------6 2.1語音的定義 -----------------------------------6 2.2 發聲原理 ------------------------------------7 2.3 何謂語音辨識 --------------------------------9 2.4 語音辨識之技術分類 --------------------------10 第三章 語音辨識演算法 -----------------------------12 3.1 語音訊號數位化 ------------------------------12 3.2 語音訊號之前置處理 --------------------------14 3.2.1 語音訊號取樣 ----------------------------14 3.2.2 移除直流偏移量 ( DC-offset removal ) ----14 3.2.3 端點偵測(Endpoint Detection) ----------15 3.2.4 音框化 ( Frame blocking )----------------19 3.2.5 預強調處理(Pre-emphasis) ----------------19 3.2.6 視窗函數(Windowing ) --------------------21 3.3 特徵參數擷取 ( Parameter Extraction ) -------22 3.3.1線性預測參數 (LPC) -----------------------23 3.3.2 梅爾倒頻參數 (MFCC) ---------------------27 3.4 動態時間較準演算法(Dynamic Time Warping, DTW)----------------------------------------31 3.5 隱藏式馬可夫模型 ( Hidden Markov Models, HMM) ----------------------------------------34 3.5.1 正算程序(Forward Procedure) -------------35 3.5.2 逆算程序(Backward Procedure) ------------37 3.5.3 維特比演算法(Viterbi Algorithm) ---------38 第四章 語者驗證系統架構 ---------------------------41 4.1 系統架構 ------------------------------------41 4.2 前端處理 ( Front-end Processing ) -----------44 4.3 特徵擷取 ( Feature extraction ) -------------50 4.4 訓練模式 ( Training Model ) -----------------53 4.5 測試架構流程 --------------------------------59 第五章 系統實驗與數據討論 -------------------------61 5.1 訓練語料的收集 ------------------------------61 5.2狀態數不同時在測試時對系統之影響 -------------63 5.3 LPC 參數模擬結果 ----------------------------66 5.4 MFCC 參數模擬結果 ---------------------------71 5.5模擬數據比較 ( LPCC, MFCC and SV ) -----------77 第六章 結論與未來展望------------------------------81 6.1結論 -----------------------------------------81 6.2未來展望 -------------------------------------82 參考文獻 ------------------------------------------84zh_TW
dc.titleA Study of Text Dependent Speaker Verificationen_US
dc.typeThesis and Dissertationzh_TW
item.openairetypeThesis and Dissertation-
item.fulltextno fulltext-
Appears in Collections:電機工程學系所
Show simple item record
TAIR Related Article

Google ScholarTM


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.