%PDF-1.4 % 5 0 obj << /S /GoTo /D (chapter*.2) >> endobj 8 0 obj (List of Tables) endobj 9 0 obj << /S /GoTo /D (chapter*.3) >> endobj 12 0 obj (List of Figures) endobj 13 0 obj << /S /GoTo /D (chapter.1) >> endobj 16 0 obj (Introduction) endobj 17 0 obj << /S /GoTo /D (section.1.1) >> endobj 20 0 obj (Speaker recognition: principles and applications) endobj 21 0 obj << /S /GoTo /D (section.1.2) >> endobj 24 0 obj (Thesis motivation and contributions) endobj 25 0 obj << /S /GoTo /D (section.1.3) >> endobj 28 0 obj (Thesis organization) endobj 29 0 obj << /S /GoTo /D (section.1.4) >> endobj 32 0 obj (Notations) endobj 33 0 obj << /S /GoTo /D (chapter.2) >> endobj 36 0 obj (Text-independent speaker identification: a brief review) endobj 37 0 obj << /S /GoTo /D (section.2.1) >> endobj 40 0 obj (Problem formulation) endobj 41 0 obj << /S /GoTo /D (section.2.2) >> endobj 44 0 obj (Feature extraction) endobj 45 0 obj << /S /GoTo /D (section.2.3) >> endobj 48 0 obj (Classification techniques) endobj 49 0 obj << /S /GoTo /D (subsection.2.3.1) >> endobj 52 0 obj (Unsupervised learning techniques) endobj 53 0 obj << /S /GoTo /D (subsection.2.3.2) >> endobj 56 0 obj (Supervised learning techniques) endobj 57 0 obj << /S /GoTo /D (section.2.4) >> endobj 60 0 obj (Mismatch reduction techniques) endobj 61 0 obj << /S /GoTo /D (subsection.2.4.1) >> endobj 64 0 obj (Feature-based compensation techniques) endobj 65 0 obj << /S /GoTo /D (subsection.2.4.2) >> endobj 68 0 obj (Model-based compensation techniques) endobj 69 0 obj << /S /GoTo /D (subsection.2.4.3) >> endobj 72 0 obj (Score-based compensation techniques) endobj 73 0 obj << /S /GoTo /D (section.2.5) >> endobj 76 0 obj (Summary and conclusions) endobj 77 0 obj << /S /GoTo /D (chapter.3) >> endobj 80 0 obj (Gaussian Mixture models) endobj 81 0 obj << /S /GoTo /D (section.3.1) >> endobj 84 0 obj (Mathematical definition of the GMM) endobj 85 0 obj << /S /GoTo /D (section.3.2) >> endobj 88 0 obj (Standard maximum likelihood framework) endobj 89 0 obj << /S /GoTo /D (subsection.3.2.1) >> endobj 92 0 obj (Parameter estimation) endobj 93 0 obj << /S /GoTo /D (subsection.3.2.2) >> endobj 96 0 obj (Classification framework) endobj 97 0 obj << /S /GoTo /D (section.3.3) >> endobj 100 0 obj (The Gaussian mixture model/universal background model framework) endobj 101 0 obj << /S /GoTo /D (chapter.4) >> endobj 104 0 obj (Vector autoregressive Gaussian Mixture model) endobj 105 0 obj << /S /GoTo /D (section.4.1) >> endobj 108 0 obj (Vector autoregressive models) endobj 109 0 obj << /S /GoTo /D (section.4.2) >> endobj 112 0 obj (Parameter estimation of the VARGM model) endobj 113 0 obj << /S /GoTo /D (subsection.4.2.1) >> endobj 116 0 obj (The general case) endobj 117 0 obj << /S /GoTo /D (subsection.4.2.2) >> endobj 120 0 obj (Diagonal autoregression matrices) endobj 121 0 obj << /S /GoTo /D (section.4.3) >> endobj 124 0 obj (Model order selection) endobj 125 0 obj << /S /GoTo /D (section.4.4) >> endobj 128 0 obj (Classification using the VARGM model) endobj 129 0 obj << /S /GoTo /D (subsection.4.4.1) >> endobj 132 0 obj (Standard VARGM/ML framework) endobj 133 0 obj << /S /GoTo /D (subsection.4.4.2) >> endobj 136 0 obj (VARGM/UBM) endobj 137 0 obj << /S /GoTo /D (chapter.5) >> endobj 140 0 obj (Generalized maximum likelihood adaptation) endobj 141 0 obj << /S /GoTo /D (section.5.1) >> endobj 144 0 obj (Main statistical model) endobj 145 0 obj << /S /GoTo /D (section.5.2) >> endobj 148 0 obj (Model parameter estimation) endobj 149 0 obj << /S /GoTo /D (section.5.3) >> endobj 152 0 obj (Selection of the optimum regression order) endobj 153 0 obj << /S /GoTo /D (section.5.4) >> endobj 156 0 obj (Adaptation using the GML rule) endobj 157 0 obj << /S /GoTo /D (section.5.5) >> endobj 160 0 obj (Blind equalization of MIMO channels) endobj 161 0 obj << /S /GoTo /D (subsection.5.5.1) >> endobj 164 0 obj (Problem formulation) endobj 165 0 obj << /S /GoTo /D (subsection.5.5.2) >> endobj 168 0 obj (Parameter estimation of the equalizer filter) endobj 169 0 obj << /S /GoTo /D (subsection.5.5.3) >> endobj 172 0 obj (The proposed equalization algorithm) endobj 173 0 obj << /S /GoTo /D (chapter.6) >> endobj 176 0 obj (Experimental Evaluation) endobj 177 0 obj << /S /GoTo /D (section.6.1) >> endobj 180 0 obj (Group I: Closed-set text-independent speaker identification using the VARGM model) endobj 181 0 obj << /S /GoTo /D (subsection.6.1.1) >> endobj 184 0 obj (The 2000 NIST speaker recognition evaluation) endobj 185 0 obj << /S /GoTo /D (subsection.6.1.2) >> endobj 188 0 obj (A comparison between GMM and VARGM) endobj 189 0 obj << /S /GoTo /D (subsection.6.1.3) >> endobj 192 0 obj (VARGM model order selection) endobj 193 0 obj << /S /GoTo /D (section.6.2) >> endobj 196 0 obj (Group II: Speech emotion recognition using the VARGM model) endobj 197 0 obj << /S /GoTo /D (subsection.6.2.1) >> endobj 200 0 obj (The Berlin emotional database) endobj 201 0 obj << /S /GoTo /D (subsection.6.2.2) >> endobj 204 0 obj (Results and discussion) endobj 205 0 obj << /S /GoTo /D (section.6.3) >> endobj 208 0 obj (Group III: Adaptive speaker identification using the GML rule) endobj 209 0 obj << /S /GoTo /D (subsection.6.3.1) >> endobj 212 0 obj (The TIMIT database) endobj 213 0 obj << /S /GoTo /D (subsection.6.3.2) >> endobj 216 0 obj (Modeling the mismatch by convolutive noise) endobj 217 0 obj << /S /GoTo /D (subsection.6.3.3) >> endobj 220 0 obj (Modeling mismatch by additive white Gaussian noise) endobj 221 0 obj << /S /GoTo /D (section.6.4) >> endobj 224 0 obj (Group IV: Blind equalization of MIMO channels) endobj 225 0 obj << /S /GoTo /D (subsection.6.4.1) >> endobj 228 0 obj (Comparison with the whitening approach) endobj 229 0 obj << /S /GoTo /D (subsection.6.4.2) >> endobj 232 0 obj (Equalization over frequency-flat slow fading channels) endobj 233 0 obj << /S /GoTo /D (subsection.6.4.3) >> endobj 236 0 obj (Separable MIMO channels) endobj 237 0 obj << /S /GoTo /D (chapter.7) >> endobj 240 0 obj (Conclusions and future work) endobj 241 0 obj << /S /GoTo /D (section.7.1) >> endobj 244 0 obj (Summary of results and thesis contribution) endobj 245 0 obj << /S /GoTo /D (section.7.2) >> endobj 248 0 obj (Future research directions) endobj 249 0 obj << /S /GoTo /D (section.7.3) >> endobj 252 0 obj (Publications) endobj 253 0 obj << /S /GoTo /D (subsection.7.3.1) >> endobj 256 0 obj (Accepted journal papers) endobj 257 0 obj << /S /GoTo /D (subsection.7.3.2) >> endobj 260 0 obj (Submitted journal papers) endobj 261 0 obj << /S /GoTo /D (subsection.7.3.3) >> endobj 264 0 obj (Accepted conference papers) endobj 265 0 obj << /S /GoTo /D (appendix*.12) >> endobj 268 0 obj (APPENDICES) endobj 269 0 obj << /S /GoTo /D (appendix.A) >> endobj 272 0 obj (Derivation of relations for the smoothed statistics in the GML framework) endobj 273 0 obj << /S /GoTo /D (appendix.B) >> endobj 276 0 obj (Proof of Theorem 1 in Chapter 5) endobj 277 0 obj << /S /GoTo /D (appendix.C) >> endobj 280 0 obj (Convergence Analysis of the EM algorithm used to estimate the equalizer filter) endobj 281 0 obj << /S /GoTo /D (appendix*.13) >> endobj 284 0 obj (Bibliography) endobj 285 0 obj << /S /GoTo /D [286 0 R /Fit ] >> endobj 288 0 obj << /Length 538 /Filter /FlateDecode >> stream xmMs0>3,J80!AJtV-ɛ;ϻڷ7頻 J+ZmvPIwڌ}:0l=4ֶi1֣LAc}j%BUDs8G{JOR`}ؚp% 9E>!%L)Y)'a|@RuEaTTJF{JX1hmեa}mF'OU;0"y@x[p#cy}_Z;{8_c'i TÃ1 \\Qr>kYm'P|ny؝ t,ޘFιcA 5LuvLΌGBY#烛N@߿.w=tDJuz8m컆x:G[^3 :l&/b㷨{0KԒ3ތ}Ye@+dr\rLځ=WO_ swϮ\C+u/ endstream endobj 286 0 obj << /Type /Page /Contents 288 0 R /Resources 287 0 R /MediaBox [0 0 612 792] /Parent 295 0 R >> endobj 289 0 obj << /D [286 0 R /XYZ 94.442 768.821 null] >> endobj 290 0 obj << /D [286 0 R /XYZ 95.442 720 null] >> endobj 287 0 obj << /Font << /F17 291 0 R /F15 292 0 R /F18 293 0 R /F19 294 0 R >> /ProcSet [ /PDF /Text ] >> endobj 298 0 obj << /Length 314 /Filter /FlateDecode >> stream xUPMO H[Rh+vZyC FHR+&[JP~clxa xGӠ3[O F0FqiLnR5*8s64f?K(H)xT۠ݕօkR GF]7e716zzcV߱=`' .KR\[9th.z+Krya،IqhǯR4jgjhyY"br9m;w-)_ endstream endobj 297 0 obj << /Type /Page /Contents 298 0 R /Resources 296 0 R /MediaBox [0 0 612 792] /Parent 295 0 R >> endobj 299 0 obj << /D [297 0 R /XYZ 94.442 768.821 null] >> endobj 296 0 obj << /Font << /F15 292 0 R >> /ProcSet [ /PDF /Text ] >> endobj 302 0 obj << /Length 1953 /Filter /FlateDecode >> stream xuXrWfUV_[n*8g1qErlf(/d[ $>OqLJ*ϓ.)CRcT,7';& a\f2/vaZ2<}Mh"c=>#Q`݃?Qj2Cpji1o, ^7[Gj'3]δvadβ dz˒adw0}8)7f:4Q>ìۉHeԼ0NxnC5llQ#T||)A72OGCPuh%buXcf|fB1%|xz&^w62slXN~9*3'lwE,?N#gWvQu= ǡ̦0 LH Wy;͍n 40}Lt!^HQMc&աJchYhMU4 c'A\.x5^&CW4xȒg<@(ъ
!bEs>lgնP 'ːw2DIM^ϰ[k:39c&8FŎlj,Z2۩ yfG0 3ڷz%z6`ư.5K W3}'q|3j;5RqF)|6rKxRabJv#kM3è&me&۰2 `*p۳n)m
xIڴ7S2
!|Ź\ޒ@4b1n:h X1V7cX};^
H?oW\"KG!49o@WAlx [\\OLeC%"D3 lR}LWg$U
[.+5_xU7LHNZu7[ܕsnQH[~\rعW.p`Jfy;0e;Q/,{"GC-|B Fա2~a_EͽTѺNc'nG$5X]$,NEnoMh[wf_Aq*/.+5Pm!q.X/ҩos'ŭ.@wJD|=I,ӘCncv9[O/8\FsΊF£D
n@gmwM7@)eNȝqMtTѺsjNըBszC9,y]2ۙO DC'ɸ>AciHz/9d_؈V|8Z*KA`\Qnl46V^Gc-^9}$[4hbҢ\r>S;OoWFpܳ&9QXi lќfyU\{7gB5pŅY
Nh:u1%!˄S})CaQj17Қ
]w拉̻XouK @2 S2ZێƯۀY\ݢGi;Nqã
?8?VӣJ+