臺灣博碩士論文加值系統

高斯混合模型GMM (Gaussian Mixture Model)係多個高斯機率密度函數的組合，它能夠平滑地逼近任意形狀的機率密度分布，在各種圖樣辨識的領域都具有不錯的辨識效果。然而在建置語者模型的過程中，其必須透過不斷地疊代計算來決定各個高斯機率密度函數的參數，運算過程相當地複雜。本研究提出改良高斯混合模型，其辨識用的各個特徵都有其獨立的高斯機率密度函數，由於不需經疊代的過程，可大幅減少計算量，且經由語者鑑定的實驗結果顯示，其仍能維持不錯的辨識效果。本研究採用梅爾倒頻譜係數作為語者鑑定用之語音特徵，傳統式高斯混合模型所比對出來的平均錯誤率為0.5901%，而改良型高斯混合模型語者鑑定的平均錯誤率則為1.6700%，其間的差距為1.0799%，兩方法之間錯誤率效果的差異性不大，但因改良型高斯混合模型在建置語者模型的過程中並不需要經過疊代運算，比傳統式高斯混合模型的運算方式和時間更加簡易與快速，更可為另一種考量速度與方便性的演算法。

Gaussian mixture model (GMM) is a combination of a plurality of Gaussian probability density function, it can be smoothly approximate the probability density distribution of any arbitrary shape. In various areas of pattern recognition, it has a good recognition results. However, during building the speaker model process, we must determine the parameters of each Gaussian probability density function through constantly iterative calculation, the calculation process is quite complex. This paper presents modified Gaussian mixture model, each characteristic for recognition has its own independent Gaussian probability density function. Since the process without iteration, it can significantly reduce the amount of calculation. And the speaker verification results show that it can still maintain a good recognition results. In this paper, we use Mel frequency cepstral coefficients(MFCCs) as the voice characteristic for speaker verification. The average error rate for speaker verification on Gaussian mixture model is 0.5901%, while it on modified Gaussian mixture model is 1.6700%, the gap between them was 1.0799%. The error rate of two methods during the speaker verification has less difference. But in the speaker model build process, modified Gaussian mixture model does not need to go through an iterative calculation. The calculation ways and time are more simple and faster than Gaussian mixture model. It can also be another consideration of algorithm for more speed and convenience.

第一章 前言 1
第一節 高斯混合模型(GMM)在各領域之應用 1
一、聲紋辨識 2
二、圖像辨識 2
三、遙測(Remote Sensing)圖像 2
四、核磁共振成像(Magnetic Resonance Imaging) 3
五、配電系統狀態估測（Distribution System State Estimation，DSSE） 3
六、半導體產業錯誤偵測(Fault detection) 3
七、智慧運輸系統(Intelligent Transportation System) 4
八、氣體辨識系統(Gas Identification Systems) 4
第二節 聲紋辨識(Voice Recognition)簡介 5
一、語音辨識(Speech Recognition) 5
二、語者辨識(Speaker Recognition) 6
三、語者識別(Speaker Identification) 6
四、語者鑑定(Speaker Verification) 7
第三節 特徵參數之類型 9
一、頻域(Frequency Domain)特徵參數 10
二、時域(Time Domain)特徵參數 11
第四節 建立語者辨識模型之演算法 15
一、統計法(Statistics) 16
二、動態時間校正(Dynamic Time Warping，DTW) 16
三、向量量化(Vector Quantization，VQ) 16
四、支援向量機( Support Vector Machine，SVM)
五、特性語音法(Eigenvoice) 17
六、隱藏馬可夫模型(Hidden Markov Model，HMM) 18
七、高斯混合模型(Gaussian Mixture Model，GMM) 19
八、類神經網路(Artificial Neural Networks) 19
第二章 梅爾頻率倒頻譜係數(MFCCs) 21
第一節 簡介 21
第二節 特徵參數擷取流程 23
一、預強調(pre-emphasis) 23
二、音框化(frame blocking) 23
三、加視窗(windowing) 24
四、端點偵測(endpoint detection) 24
五、快速傅立葉轉換(Fast Fourier Transform，FFT) 25
六、三角帶通濾波器(triangular band-pass filters) 25
七、離散餘弦轉換(Discrete Cosine Transform，DCT) 28
第三節 實驗材料 30
一、語料庫 30
二、實驗材料相關參數 31
第三章 高斯混合模型(GMM) 33
第一節 高斯模型簡介 33
第二節 單一高斯機率密度函數 34
第三節 高斯混合密度函數 35
一、樣本分類已知情況下的高斯混合模型 37
二、樣本分類未知情況下的高斯混合模型 37
第四節 簡例推導過程 41
第四章 研究方法 45
第一節 簡介 45
第二節 改良型高斯混合模型 46
一、計算各維度之平均值 46
二、計算各維度之標準差 46
三、建立已知語音之相似度模型 47
第三節 簡例推導過程 49
第四節 待鑑定語音特徵比對 50
第五節 系統錯誤率估算 51
一、非本人樣本相似度分數的常態分佈 51
二、本人樣本相似度分數的常態分佈 52
三、錯誤率評估 53
第五章 實驗過程與結果 57
第一節 實驗前提 57
一、實驗目的 57
二、實驗方法 57
第二節 相似度比對結果 58
一、高斯混合模型之相似度分數 58
二、改良型高斯混合模型之相似度分數 73
第三節 比對錯誤率結果 88
一、以高斯混合模型比對之錯誤率 88
二、以改良型高斯混合模型比對之錯誤率 98
三、平均錯誤率 108
第四節 結果討論 110
一、演算過程 110
二、複雜度計算 112
三、錯誤率討論 113
第六章 結論與未來研究 115
參考文獻 117

參考文獻
1.P. McKenzie and M. Alder, "Initializing the EM algorithm for use in Gaussian mixture modelling," in Machine Intelligence and Pattern Recognition. vol. Volume 16, S. G. Edzard and S. K. Laveen, Eds., ed: North-Holland, 1994, pp. 91-105.
2.B. L. Pellom and J. H. L. Hansen, "An efficient scoring algorithm for Gaussian mixture model based speaker identification," Signal Processing Letters, IEEE, vol. 5, pp. 281-284, 1998.
3.V. V. Digalakis, D. Rtischev, and L. G. Neumeyer, "Speaker adaptation using constrained estimation of Gaussian mixtures," Speech and Audio Processing, IEEE Transactions on, vol. 3, pp. 357-366, 1995.
4.D. A. Reynolds, "Speaker identification and verification using Gaussian mixture speaker models," Speech Communication, vol. 17, pp. 91-108, 8// 1995.
5.A. Dempster, N. Laird, and D. Rubin, "MaximumL ikelihoodf romI ncompleteD ata via the EM Algorithm," Journal of the Royal Statistical Society, vol. 39, pp. 1-38, 1977.
6.X. Bing and T. Berger, "Efficient text-independent speaker verification with structural Gaussian mixture models and neural network," Speech and Audio Processing, IEEE Transactions on, vol. 11, pp. 447-456, 2003.
7.H. Zhenkun, H. Ruimin, and W. Zhongyuan, "Background Subtraction With Video Coding," Signal Processing Letters, IEEE, vol. 20, pp. 1058-1061, 2013.
8.Y. Hua and Z. Xiao-Ping, "Statistical Modeling in the Wavelet Domain for Compact Feature Extraction and Similarity Measure of Images," Circuits and Systems for Video Technology, IEEE Transactions on, vol. 20, pp. 439-445, 2010.
9.K. Nandakumar, C. Yi, S. C. Dass, and A. K. Jain, "Likelihood Ratio-Based Biometric Score Fusion," Pattern Analysis and Machine Intelligence, IEEE Transactions on, vol. 30, pp. 342-347, 2008.
10.D. Mukherjee, Q. M. J. Wu, and N. Thanh Minh, "Gaussian Mixture Model With Advanced Distance Measure Based on Support Weights and Histogram of Gradients for Background Suppression," Industrial Informatics, IEEE Transactions on, vol. 10, pp. 1086-1096, 2014.
11.L. Tengfei, J. Weili, H. Guojin, and W. Wei, "Automatic Line Segment Registration Using Gaussian Mixture Model and Expectation-Maximization Algorithm," Selected Topics in Applied Earth Observations and Remote Sensing, IEEE Journal of, vol. 7, pp. 1688-1699, 2014.
12.P. Scheunders and S. De Backer, "Wavelet Denoising of Multicomponent Images Using Gaussian Scale Mixture Models and a Noise-Free Image as Priors," Image Processing, IEEE Transactions on, vol. 16, pp. 1865-1872, 2007.
13.T. GuangJian, X. Yong, Y. Zhang, and F. Dagan, "Hybrid Genetic and Variational Expectation-Maximization Algorithm for Gaussian-Mixture-Model-Based Brain MR Image Segmentation," Information Technology in Biomedicine, IEEE Transactions on, vol. 15, pp. 373-380, 2011.
14.H. Greenspan, A. Ruf, and J. Goldberger, "Constrained Gaussian mixture model framework for automatic segmentation of MR brain images," Medical Imaging, IEEE Transactions on, vol. 25, pp. 1233-1245, 2006.
15.L. Xiaoyun and C. Fen, "Automatic Segmentation of 3-D Brain MR Images by Using Global Tissue Spatial Structure Information," Applied Superconductivity, IEEE Transactions on, vol. 24, pp. 1-5, 2014.
16.E. Manitsas, R. Singh, B. C. Pal, and G. Strbac, "Distribution System State Estimation Using an Artificial Neural Network Approach for Pseudo Measurement Modeling," Power Systems, IEEE Transactions on, vol. 27, pp. 1888-1896, 2012.
17.Y. Jianbo, "Fault Detection Using Principal Components-Based Gaussian Mixture Model for Semiconductor Manufacturing Processes," Semiconductor Manufacturing, IEEE Transactions on, vol. 24, pp. 432-444, 2011.
18.V. A. Butakov and P. A. Ioannou, "Driver/Vehicle Response Diagnostic System for the Vehicle-Following Case," Intelligent Transportation Systems, IEEE Transactions on, vol. 15, pp. 1947-1957, 2014.
19.S. Brahim-Belhouari, A. Bermak, S. Minghua, and P. C. H. Chan, "Fast and robust gas identification system using an integrated gas sensor technology and Gaussian mixture models," Sensors Journal, IEEE, vol. 5, pp. 1433-1444, 2005.
20.A. Fazel and S. Chakrabartty, "An Overview of Statistical Pattern Recognition Techniques for Speaker Verification," Circuits and Systems Magazine, IEEE, vol. 11, pp. 62-81, 2011.
21.J. W. Picone, "Signal modeling techniques in speech recognition," Proceedings of the IEEE, vol. 81, pp. 1215-1247, 1993.
22.P. Tsang-Long, L. Wen-Yuan, W. Tsan-Nung, and L. Ching-Yi, "Automatic visual feature extraction for Mandarin audio-visual speech recognition," in Systems, Man and Cybernetics, 2009. SMC 2009. IEEE International Conference on, 2009, pp. 2936-2940.
23.M. N. Kaynak, Z. Qi, A. D. Cheok, K. Sengupta, J. Zhang, and C. Ko, "Analysis of lip geometric features for audio-visual speech recognition," Systems, Man and Cybernetics, Part A: Systems and Humans, IEEE Transactions on, vol. 34, pp. 564-570, 2004.
24.N. Singh, R. A. Khan, and R. Shree, "Applications of Speaker Recognition," Procedia Engineering, vol. 38, pp. 3122-3126, // 2012.
25.D. A. Reynolds, "Large population speaker identification using clean and telephone speech," Signal Processing Letters, IEEE, vol. 2, pp. 46-48, 1995.
26.T. Kinnunen, E. Karpov, and P. Franti, "Real-time speaker identification and verification," Audio, Speech, and Language Processing, IEEE Transactions on, vol. 14, pp. 277-288, 2006.
27.T. G. Clarkson, C. C. Christodoulou, G. Yelin, D. Gorse, D. A. Romano-Critchley, and J. G. Taylor, "Speaker identification for security systems using reinforcement-trained pRAM neural network architectures," Systems, Man, and Cybernetics, Part C: Applications and Reviews, IEEE Transactions on, vol. 31, pp. 65-76, 2001.
28.J. P. Campbell, W. Shen, W. M. Campbell, R. Schwartz, J. F. Bonastre, and D. Matrouf, "Forensic speaker recognition," Signal Processing Magazine, IEEE, vol. 26, pp. 95-103, 2009.
29.王柏強，基於梅爾倒頻譜係數語者鑑定之參數選擇，中央警察大學刑事警察研究所碩士論文，民國104年。
30.W. Ning, P. C. Ching, Z. Nengheng, and L. Tan, "Robust Speaker Recognition Using Denoised Vocal Source and Vocal Tract Features," Audio, Speech, and Language Processing, IEEE Transactions on, vol. 19, pp. 196-205, 2011.
31.D. Burton, "Text-dependent speaker verification using vector quantization source coding," Acoustics, Speech and Signal Processing, IEEE Transactions on, vol. 35, pp. 133-143, 1987.
32.A. M. T. S. B. Adikari, S. Devadithya, A. R. S. T. Bandara, K. C. J. Dharmawardane, and K. C. B. Wavegedara, "Application of automatic speaker verification techniques for forensic evidence evaluation," in Digital Signal Processing (DSP), 2014 19th International Conference on, 2014, pp. 444-448.
33.H. Gish and M. Schmidt, "Text-independent speaker identification," Signal Processing Magazine, IEEE, vol. 11, pp. 18-32, 1994.
34.J. Ortega-Garcia, J. Gonzalez-Rodriguez, and S. Cruz-Llanas, "Speech variability in automatic speaker recognition systems for commercial and forensic purposes," Aerospace and Electronic Systems Magazine, IEEE, vol. 15, pp. 27-32, 2000.
35.T. Kinnunen and H. Li, "An overview of text-independent speaker recognition: From features to supervectors," Speech Communication, vol. 52, pp. 12-40, 1// 2010.
36.W. Hess, "A pitch-synchronous digital feature extraction system for phonemic recognition of speech," Acoustics, Speech and Signal Processing, IEEE Transactions on, vol. 24, pp. 14-25, 1976.
37.S. Furui, "Comparison of speaker recognition methods using statistical features and dynamic features," Acoustics, Speech and Signal Processing, IEEE Transactions on, vol. 29, pp. 342-350, 1981.
38.J. Pohjalainen, C. Hanilci, T. Kinnunen, and P. Alku, "Mixture Linear Prediction in Speaker Verification Under Vocal Effort Mismatch," Signal Processing Letters, IEEE, vol. 21, pp. 1516-1520, 2014.
39.M. D. Plumpe, T. F. Quatieri, and D. A. Reynolds, "Modeling of the glottal flow derivative waveform with application to speaker identification," Speech and Audio Processing, IEEE Transactions on, vol. 7, pp. 569-586, 1999.
40.K. T. Assaleh and R. J. Mammone, "New LP-derived features for speaker identification," Speech and Audio Processing, IEEE Transactions on, vol. 2, pp. 630-638, 1994.
41.M. Farrus and J. Hernando, "Using jitter and shimmer in speaker verification," Signal Processing, IET, vol. 3, pp. 247-257, 2009.
42.J. Huang and Z. Yunxin, "Energy-constrained signal subspace method for speech enhancement and recognition," Signal Processing Letters, IEEE, vol. 4, pp. 283-285, 1997.
43.L. Yiu-Kei and C. Chok-Ki, "Speech recognition based on zero crossing rate and energy," Acoustics, Speech and Signal Processing, IEEE Transactions on, vol. 33, pp. 320-323, 1985.
44.B. S. Atal and L. Rabiner, "A pattern recognition approach to voiced-unvoiced-silence classification with applications to speech recognition," Acoustics, Speech and Signal Processing, IEEE Transactions on, vol. 24, pp. 201-212, 1976.
45.Q. Ibrahim and N. Abdulghani, "Security enhancement of voice over Internet protocol using speaker recognition technique," Communications, IET, vol. 6, pp. 604-612, 2012.
46.B. Bhanu and J. H. McClellan, "On the computation of the complex cepstrum," Acoustics, Speech and Signal Processing, IEEE Transactions on, vol. 28, pp. 583-585, 1980.
47.B. Strope and A. Alwan, "A model of dynamic auditory perception and its application to robust word recognition," Speech and Audio Processing, IEEE Transactions on, vol. 5, pp. 451-464, 1997.
48.P. Delsarte and Y. V. Genin, "The split Levinson algorithm," Acoustics, Speech and Signal Processing, IEEE Transactions on, vol. 34, pp. 470-478, 1986.
49.黃中珉，基於梅爾倒頻譜之語者鑑定，中央警察大學刑事警察研究所碩士論文，民國102年。
50.M. Sambur and L. Rabiner, "A statical decision approach to the recognition of connected digits," Acoustics, Speech and Signal Processing, IEEE Transactions on, vol. 24, pp. 550-558, 1976.
51.詹明華，刑事圖樣鑑識(一)，中央警察大學出版社，民國90年。
52.A. D. Subramaniam, W. R. Gardner, and B. D. Rao, "Low-complexity source coding using Gaussian mixture models, lattice vector quantization, and recursive coding with application to speech spectrum quantization," Audio, Speech, and Language Processing, IEEE Transactions on, vol. 14, pp. 524-532, 2006.
53.S. Cumani and P. Laface, "Analysis of Large-Scale SVM Training Algorithms for Language and Speaker Recognition," Audio, Speech, and Language Processing, IEEE Transactions on, vol. 20, pp. 1585-1596, 2012.
54.P. Kenny, G. Boulianne, and P. Dumouchel, "Eigenvoice modeling with sparse training data," Speech and Audio Processing, IEEE Transactions on, vol. 13, pp. 345-354, 2005.
55.L. Rabiner, "A tutorial on hidden Markov models and selected applications in speech recognition," Proceedings of the IEEE, vol. 77, pp. 257-286, 1989.
56.N. Morgan and H. A. Bourlard, "Neural networks for statistical recognition of continuous speech," Proceedings of the IEEE, vol. 83, pp. 742-772, 1995.
57.E. Trentin and M. Gori, "Robust combination of neural networks and hidden Markov models for speech recognition," Neural Networks, IEEE Transactions on, vol. 14, pp. 1519-1531, 2003.
58.D. G. Childers, D. P. Skinner, and R. C. Kemerait, "The cepstrum: A guide to processing," Proceedings of the IEEE, vol. 65, pp. 1428-1443, 1977.
59.Z. Qifeng and A. Alwan, "On the use of variable frame rate analysis in speech recognition," in Acoustics, Speech, and Signal Processing, 2000. ICASSP '00. Proceedings. 2000 IEEE International Conference on, 2000, pp. 1783-1786 vol.3.
60.Z. Yiying, D. Zhang, and Z. Xiaoyan, "A novel text-independent speaker verification method based on the global speaker model," Systems, Man and Cybernetics, Part A: Systems and Humans, IEEE Transactions on, vol. 30, pp. 598-602, 2000.
61.G. Liuyang, G. Yunfei, L. Shaomei, and C. Fucai, "Speech enhancement algorithm based on improved spectral subtraction," in Intelligent Computing and Intelligent Systems, 2009. ICIS 2009. IEEE International Conference on, 2009, pp. 140-143.
62.S. Nakagawa, W. Longbiao, and S. Ohtsuka, "Speaker Identification and Verification by Combining MFCC and Phase Information," Audio, Speech, and Language Processing, IEEE Transactions on, vol. 20, pp. 1085-1095, 2012.
63.A. K. Sarkar, D. Cong-Thanh, L. Viet-Bac, and C. Barras, "Combination of Cepstral and Phonetically Discriminative Features for Speaker Verification," Signal Processing Letters, IEEE, vol. 21, pp. 1040-1044, 2014.
64.Z. Tychtl and J. Psutka, "Speech production based on the mel-frequency cepstral coefficients," in EuroSpeech, 1999, pp. 2335-2338.
65.MAT Speech Database -TCC300Edu [Online]，引用；擷取自：http://www.aclclp.org.tw/use_mat.php#tcc300edu
66.吳晉瑋，基於梅爾倒頻譜之線上語者鑑定，中央警察大學刑事警察研究所碩士論文，民國103年。
67.R. Bolle, Guide to Biometrics: Springer, 2004.