| { | |
| "paper_id": "O14-1004", | |
| "header": { | |
| "generated_with": "S2ORC 1.0.0", | |
| "date_generated": "2023-01-19T08:04:25.188635Z" | |
| }, | |
| "title": "Sparse Representation Based Speaker Identification", | |
| "authors": [ | |
| { | |
| "first": "Kuang-Yao", | |
| "middle": [], | |
| "last": "\u738b\u5149\u8000", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "National Central University", | |
| "location": {} | |
| }, | |
| "email": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Wang", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "National Central University", | |
| "location": {} | |
| }, | |
| "email": "" | |
| }, | |
| { | |
| "first": "Jia-Ching", | |
| "middle": [], | |
| "last": "\u738b\u5bb6\u6176", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "National Central University", | |
| "location": {} | |
| }, | |
| "email": "" | |
| } | |
| ], | |
| "year": "", | |
| "venue": null, | |
| "identifiers": {}, | |
| "abstract": "", | |
| "pdf_parse": { | |
| "paper_id": "O14-1004", | |
| "_pdf_hash": "", | |
| "abstract": [], | |
| "body_text": [ | |
| { | |
| "text": "\u7b2c\u4e00\u90e8\u5206\u70ba\u7c21\u4ecb\uff0c\u7b2c\u4e8c\u90e8\u5206\u70ba\u53c3\u6578\u64f7\u53d6\uff0c\u6211\u5011\u4ee5 PPCA \u5efa\u69cb\u8d85\u7d1a\u5411\u91cf(PPCA \u8d85\u7d1a\u5411\u91cf) \u53d6\u4ee3 GMM \u8d85\u7d1a\u5411\u91cf[20][31]\uff0c\u4e26\u4e14\u4ee5\u5df4\u96f7\u7279\u6aa2\u5b9a(Bartlett Test)\u7684\u65b9\u5f0f\u8abf\u6574\u7279\u5fb5\u503c\u7684\u9078 \u53d6\uff0c\u7b2c\u4e09\u90e8\u5206\u5247\u662f\u63cf\u8ff0\u5982\u4f55\u5229\u7528\u7279\u5fb5\u53c3\u6578\u5efa\u7acb\u7a00\u758f\u8868\u793a\u5206\u985e\u5668\u3002\u7b2c\u56db\u90e8\u5206\u63cf\u8ff0\u6211\u5011\u63d0\u51fa \u7684 \u5169 \u7a2e \u8b8a \u7570 \u88dc \u511f \u65b9 \u6cd5 \uff0c \u4f4e \u79e9 \u77e9 \u9663 \u9084 \u539f \u4ee5 \u53ca \u6838 \u5316 \u7a00 \u758f \u8868 \u793a \u5206 \u985e \u5668 (Kernel", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "(1) ", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u5176\u4e2d x \u4ee3\u8868\u9ad8\u7dad\u7684\u8cc7\u6599\uff0cz \u4ee3\u8868\u4f4e\u7dad\u8b8a\u6578\uff0c\u53c8\u7a31 Latent Factor\u3002\u56de\u60f3\u4e3b\u6210\u5206\u5206\u6790\u7684\u6578\u5b78 \u5f0f\uff0c\u627e\u51fa\u4e3b\u6210\u5206 V \u4f7f\u8cc7\u6599\u964d\u7dad\uff1a Vz \u03bc x \uf02b \uf03d (2) \u6211\u5011\u53ef\u5c07\u6210\u4efd\u5206\u6790\u8996\u6210\u5c0d\u8cc7\u6599 x \u505a PCA \u52a0\u4e0a\u4e00\u500b\u566a\u97f3\u9805\uff0c\u4e26\u4e14\u5c0e\u5165\u6a5f\u7387\u7684\u6982\u5ff5\u89e3\u91cb\u3002 \u5728\u5f0f(1)\u4e2d\uff0cx \u662f 1 \uf0b4 K \u7684\u539f\u59cb\u8cc7\u6599\uff0c W \u70ba J K \uf0b4 \u7684\u8f49\u63db\u77e9\u9663\uff0c\u5176\u4e2d K J \uf03c \uff0c\u800c Latent Factor z \u5047\u8a2d\u70ba\u4e00\u9ad8\u65af\u5206\u5e03 ) I , 0 ( N \uff0c\u566a\u97f3 ) I , 0 ( 2 \uf073 N \u03b5 \uff0c\u6839\u64da\u4ee5\u4e0a\u7684\u5047\u8a2d\u53ef\u4ee5\u77e5\u9053\u539f\u59cb\u8cc7 \u6599 x \u80fd\u5efa\u6a21(Model)\u6210 ) , ( T 2 WW I \u03bc \uf02b \uf073 N \uff0c\u7a31\u4f5c PPCA \u6a21\u578b\u3002\u7576\u9047\u5230\u66f4\u8907\u96dc\u7684\u8cc7\u6599\u6642\uff0c\u5e0c \u671b\u85c9\u7531\u6df7\u5408\u591a\u7d44 PPCA \u6a21\u578b\uff0c\u5982\u4e0b\u5f0f\uff1a \uf0e5 \uf03d \uf03d M c c c p w p 1 ) | ( ) ( x x (3) ) , ( ) | ( T 2 c c c c N c p W W I \u03bc x \uf02b \uf03d \uf073 (4) \u5176\u4e2d M \u662f\u9ad8\u65af\u6210\u4efd\u7684\u500b\u6578\uff0c\u9019\u53ef\u4ee5\u8ddf\u9ad8\u65af\u6df7\u5408\u6a21\u578b\u505a\u5c0d\u61c9\uff0c\u4e00\u822c\u8868\u793a\u8cc7\u6599\u4ee5\u591a\u500b\u9ad8\u65af \u6a21\u578b\u63cf\u8ff0\uff0c\u9019\u88e1\u7684\u57fa\u672c\u55ae\u4f4d\u4ee5 PPCA \u6a21\u578b\u53d6\u4ee3\u3002 \u56e0\u6b64\uff0c\u6211\u5011\u5e0c\u671b\u627e\u51fa c W \u53ca c \uf073 \u4f86\u63a8\u7b97 Latent Factor\uff0c\u65b9\u6cd5\u662f\u5229\u7528\u6700\u5927\u6982\u4f3c\u4f30\u8a08 (Maximum Likelihood Estimation, MLE)\uff0c\u56e0\u70ba\u5df2\u7d93\u77e5\u9053 x \u7684\u5206\u5e03\uff0c\u6240\u4ee5\u85c9\u7531 MLE \u5f97\u5230 \u7684 c W \u53ca c \uf073 \u53ef\u9032\u800c\u63a8\u51fa Latent Factor c z ) ( ) ( 1 2 c T c c c T c c \u03bc x W I W W z \uf02d \uf02b \uf03d \uf02d \uf073 (5) \u7576\u539f\u59cb\u8cc7\u6599 x \u9032\u5165\u7cfb\u7d71\u5f8c\uff0c\u6703\u5148\u4ee5 PPCA \u6df7\u5408\u6a21\u578b\uff0c\u4e26\u4e14\u900f\u904e\u5f0f(5)\u7684\u8f49\u63db\u5f97\u5230\u5c6c\u65bc \u6bcf\u500b\u9ad8\u65af\u6210\u4efd\u7684 Latent Factor\u3002\u7576\u6240\u6709\u7684\u53c3\u6578\u90fd\u505a\u4e86 PPCA \u5f8c\uff0c\u5247\u539f\u59cb\u4ee5 MFCC \u70ba\u57fa\u790e \u7684 UBM \u4e5f\u5fc5\u9808\u8abf\u6574\uff0c\u5f62\u6210\u4ee5 Latent Factor \u70ba\u53c3\u6578\u4e0b\u69cb\u5efa\u7684\u65b0 UBM\uff1a \uf0e5 \uf0e5 \uf03d \uf02d \uf03d \uf02d \uf03d \uf03d M c c c c M c c z c z c N w N w c p 1 1 2 1 , , ) , 0 ( ) , ( ) | ( \u039b I \u03a3 \u03bc z \uf073 (6) \u5176\u4e2d 0 )} ( ) {( 1 2 , \uf03d \uf02d \uf02b \uf03d \uf02d c T c c c T c c z E \u03bc x W I W W \u03bc \uf073 (7) 1 2 2 / 1 2 2 / 2 1 2 1 2 , , , ) ( ) ( ) ( } ) )( {( ) ( } { \uf02d \uf02d \uf02d \uf02d \uf02d \uf02d \uf03d \uf02d \uf02d \uf03d \uf02b \uf02d \uf02d \uf02b \uf03d \uf02d \uf03d c c T c c c c T c c c T c c T c c T c c T c c c T c T c z c z T c c c z E E \u039b I \u039b I \u039b \u039b I \u039b \u039b I W W W \u03bc x \u03bc x W I W W \u03bc \u03bc z z \u03a3 \uf073 \uf073 \uf073 \uf073 \uf073 (8) 2.2 \u5df4\u96f7\u7279\u6aa2\u5b9a \u5728\u4e0a\u500b\u5c0f\u7bc0\u4e2d\u63d0\u5230\u7684 PPCA \u4e2d\uff0c\u6709\u4e00\u500b\u554f\u984c\u662f\u9700\u8981\u8a0e\u8ad6\u7684\uff0c\u90a3\u5c31\u662f\u4e3b\u8ef8\u500b\u6578\u7684\u6311 \u9078\u3002\u6211\u5011\u5047\u8a2d\u5f8c\u9762\u4e0d\u8981\u7684\u7279\u5fb5\u503c\uff0c\u5176\u6578\u503c\u5c0f\u5230\u8db3\u4ee5\u8996\u70ba\u76f8\u540c\uff0c\u56e0\u6b64\uff0c\u5b9a\u7fa9\u5047\u8aaa n K k H \uf06c \uf06c \uf06c \uf03d \uf03d \uf03d \uf02b \uf02b ... : 2 1 0 \uff0c\u5e0c\u671b\u8ff4\u5708\u7531 n-1 \u5f80\u524d\u627e\u5230\u9580\u6abb\u503c k\u3002\u800c\u5982\u679c\u6211\u5011\u5c07\u7279\u5fb5\u503c\u7684\u5927 \u5c0f\u8996\u70ba\u9ad8\u65af\u5206\u5e03\uff0c\u5247\u5df4\u96f7\u7279\u6aa2\u5b9a[22][23]\u53ef\u4ee5\u6574\u7406\u6210\u76f8\u4f3c\u5ea6\u7684\u6aa2\u9a57\uff0c\u5982\u4e0b\u5f0f\uff1a k n n k q q n k q q k n T \uf02d \uf02b \uf03d \uf02b \uf03d \uf0f7 \uf0f7 \uf0f7 \uf0f7 \uf0f8 \uf0f6 \uf0e7 \uf0e7 \uf0e7 \uf0e7 \uf0e8 \uf0e6 \uf02d \uf03d \uf0e5 \uf0d5 1 1 \uf06c \uf06c (9) \u7576\u89c0\u6e2c\u8cc7\u6599\u6578 m \u5920\u591a\u7684\u8a71\uff0c\u53ef\u5c07T \u8996\u70ba\u4e00\u500b 2 X \u5206\u5e03\uff0c\u4e14\u5c07\u4e0a\u5f0f\u8fd1\u4f3c\u65bc\u4e0b\u5f0f\uff1a \uf028 \uf029 \uf0f7 \uf0f7 \uf0f8 \uf0f6 \uf0e7 \uf0e7 \uf0e8 \uf0e6 \uf02d \uf02d \uf02d \uf02d \uf0bb \uf0e5 \uf02b \uf03d n k q q k n n m T 1 log log ) ( 6 / ) 11 2 ( \uf06c \uf06c (10) \u5176\u4e2d \uf06c \u662f\u5f8c n-k \u500b\u7279\u5fb5\u503c\u7684\u5e73\u5747\uff0c\u5728\u6aa2\u5b9a\u904e\u7a0b\u4e2d\uff0c\u7576 n X T , 2 \uf061 \uf03e \u6642\uff0c\u5247\u5426\u6c7a\u5047\u8aaa 0 H \uff0c\u6aa2 \u5b9a\u7d42\u6b62\uff0cq \u5373\u70ba\u6211\u5011\u6311\u9078\u7684\u4e3b\u8ef8\u500b\u6578\uff0c\u5176\u4e2d\uf061 \u70ba 2 X", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u8a9e\u8005\u985e\u5225\u7684\u5b57\u5178\u7d44\u5728\u4e00\u8d77 Q i i i y 1 } , { \uf03d a \uff0c\u5176\u4e2d a i \uf0ce P , } ,..., 2 , 1 { c y i \uf0ce , \u4e00\u7b46\u6e2c\u8a66\u8cc7\u6599 a\uf0ce P . \u4ee4 \u03a6 \u8868\u793a ) , ( \uf0d7 \uf0d7 k \u7684\u975e\u7dda\u6027\u6620\u5c04\uff0c\u53ef\u5c07\u8f38\u5165\u7a7a\u9593 Input Space \uf063 \u7684\u8cc7\u6599\u6295\u5c04\u5230\u9ad8\u7dad\u7a7a \u9593 F : F \uf0ce \uf0ae \uf0ce \uf046 ) \u03a6( : a a \uf063 (17) \u8207\u7a00\u758f\u8868\u793a\u5206\u985e\u5668\u76f8\u540c\uff0c\u53c3\u6578\u7a7a\u9593\u7684\u6e2c\u8a66\u8cc7\u6599\u53ef\u8868\u793a\u70ba\u8a13\u7df4\u8cc7\u6599\u7684\u7dda\u6027\u7d44\u5408\uff1a x \u03a6 a a A \uf03d \uf046 \uf03d \uf046 \uf0e5 \uf03d ) ( ) ( 1 i Q i i x (18) \u5176\u4e2d \uf05b \uf05d ) \u03a6( ),..., \u03a6( ), \u03a6( 2 1 Q a a a \u03a6 A \uf03d \uff0c\u4e14 x = [x 1 , x 2 , \u2026, x Q ] T \u3002 \u6839\u64da\u7a00\u758f\u8868\u793a\u7684\u6982\u5ff5\uff0c\u7a00\u758f\u4fc2\u6578\u5411\u91cf x \u53ef\u4ee5\u5f9e\u4e0b\u5f0f\u6700\u4f73\u5316\u554f\u984c\u89e3\u51fa\uff1a \uf065 \uf0a3 \uf02d \uf03d 2 1 ) \u03a6( o subject t min ar\u011d x \u03a6 a x x A x (19) \u7576\u6838\u5316\u7a7a\u9593\u7684\u7dad\u5ea6\u662f\u672a\u77e5\u4e14\u9060\u9ad8\u65bc\u8a13\u7df4\u8cc7\u6599\u7684\u500b\u6578\u6642\uff0c\u6703\u5c0e\u81f4\u5f0f\u5b50(14)\u7684\u89e3\u6703\u4e0d\u5920\u7a00 \u758f\uff0c\u56e0\u6b64\u6703\u9700\u8981\u5728\u53c3\u6578\u7a7a\u9593\u9032\u884c\u964d\u4f4e\u7dad\u5ea6\u7684\u52d5\u4f5c\u3002\u4ee4 P \u8868\u793a\u6295\u5c04\u77e9\u9663 (Projection Matrix)\u3002\u5247(19)\u53ef\u6539\u6210\u4e0b\u5f0f\uff1a \uf065 \uf0a3 \uf02d \uf03d 2 1 ) \u03a6( o subject t min ar\u011d x \u03a6 P a P x x A x T T (20) \u5177\u6709\u6700\u5c0f\u91cd\u5efa\u8aa4\u5dee(Reconstruction Residual)\u7684\u985e\u5225\u5247\u70ba\u6838\u5316\u7a00\u758f\u8868\u793a\u5206\u985e\u5668\u7684\u5206\u985e\u7d50 \u679c\uff1a ) ( ) \u03a6( ) ( min ar\u011d i T T i i r i x \u03a6 P a P a A \uf064 \uf02d \uf03d \uf03d (21) \u5176\u4e2d ) ( i x \uf064 \u662f x \u7684\u7b2c i \u500b\u8a9e\u8005\u7684\u975e\u96f6\u7a00\u758f\u7cfb\u6578\u3002 Kernel", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| } | |
| ], | |
| "back_matter": [], | |
| "bib_entries": { | |
| "BIBREF0": { | |
| "ref_id": "b0", | |
| "title": "Robust text-independent speaker identification using Gaussian mixture models", | |
| "authors": [ | |
| { | |
| "first": "D", | |
| "middle": [ | |
| "A" | |
| ], | |
| "last": "Reynolds", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "R", | |
| "middle": [ | |
| "C" | |
| ], | |
| "last": "Rose", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1995, | |
| "venue": "IEEE Trans. Speech Audio Process", | |
| "volume": "3", | |
| "issue": "1", | |
| "pages": "72--83", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "D. A. Reynolds and R. C. Rose, \"Robust text-independent speaker identification using Gaussian mixture models,\" IEEE Trans. Speech Audio Process., vol. 3, no. 1, pp. 72-83, Jan. 1995.", | |
| "links": null | |
| }, | |
| "BIBREF1": { | |
| "ref_id": "b1", | |
| "title": "An efficient scoring algorithm for Gaussian mixture model based speaker identification", | |
| "authors": [ | |
| { | |
| "first": "B", | |
| "middle": [ | |
| "L" | |
| ], | |
| "last": "Pellom", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J", | |
| "middle": [ | |
| "H L" | |
| ], | |
| "last": "Hansen", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1998, | |
| "venue": "IEEE Signal Process. Lett", | |
| "volume": "5", | |
| "issue": "11", | |
| "pages": "281--284", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "B. L. Pellom and J. H. L. Hansen, \"An efficient scoring algorithm for Gaussian mixture model based speaker identification,\" IEEE Signal Process. Lett., vol. 5, no. 11, pp. 281-284, Nov. 1998.", | |
| "links": null | |
| }, | |
| "BIBREF2": { | |
| "ref_id": "b2", | |
| "title": "Support vector machines for speaker and language recognition", | |
| "authors": [ | |
| { | |
| "first": "W", | |
| "middle": [ | |
| "M" | |
| ], | |
| "last": "Campbell", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J", | |
| "middle": [ | |
| "P" | |
| ], | |
| "last": "Campbell", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "D", | |
| "middle": [ | |
| "A" | |
| ], | |
| "last": "Reynolds", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "E", | |
| "middle": [], | |
| "last": "Singer", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "P", | |
| "middle": [ | |
| "A" | |
| ], | |
| "last": "Torres-Carrasquillo", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2006, | |
| "venue": "Comput. Speech Lang", | |
| "volume": "20", | |
| "issue": "", | |
| "pages": "210--229", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "W. M. Campbell, J. P. Campbell, D. A. Reynolds, E. Singer, and P. A. Torres-Carrasquillo, \"Support vector machines for speaker and language recognition,\" Comput. Speech Lang., vol. 20, pp. 210-229, 2006.", | |
| "links": null | |
| }, | |
| "BIBREF3": { | |
| "ref_id": "b3", | |
| "title": "Maximum a posteriori estimation for multivariate Gaussian mixture observations of Markov chains", | |
| "authors": [ | |
| { | |
| "first": "J", | |
| "middle": [ | |
| "L" | |
| ], | |
| "last": "Gauvain", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "C", | |
| "middle": [ | |
| "H" | |
| ], | |
| "last": "Lee", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1994, | |
| "venue": "IEEE Trans. Speech Audio Process", | |
| "volume": "2", | |
| "issue": "2", | |
| "pages": "291--298", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "J. L. Gauvain and C. H. Lee, \"Maximum a posteriori estimation for multivariate Gaussian mixture observations of Markov chains,\" IEEE Trans. Speech Audio Process., vol. 2, no. 2, pp. 291-298, 1994.", | |
| "links": null | |
| }, | |
| "BIBREF4": { | |
| "ref_id": "b4", | |
| "title": "Speaker identification using Mel frequency cepstral coefficients", | |
| "authors": [ | |
| { | |
| "first": "M", | |
| "middle": [ | |
| "R" | |
| ], | |
| "last": "Hasan", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Jamil", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [ | |
| "G" | |
| ], | |
| "last": "Rabbani", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [ | |
| "S" | |
| ], | |
| "last": "Rahman", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2004, | |
| "venue": "3rd international Conference on Electrical & Computer Engineering ICECE", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "M. R. Hasan, M. Jamil, M. G. Rabbani, and M. S. Rahman, \"Speaker identification using Mel frequency cepstral coefficients,\" 3rd international Conference on Electrical & Computer Engineering ICECE 2004, 28-30 December 2004, Dhaka, Bangladesh.", | |
| "links": null | |
| }, | |
| "BIBREF5": { | |
| "ref_id": "b5", | |
| "title": "Perceptual Linear Predictive (PLP) Analysis of Speech", | |
| "authors": [ | |
| { | |
| "first": "H", | |
| "middle": [], | |
| "last": "Hermansky", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1990, | |
| "venue": "Journal of the Acoust. Society ofAmer", | |
| "volume": "87", | |
| "issue": "", | |
| "pages": "1738--1752", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "H. Hermansky. \"Perceptual Linear Predictive (PLP) Analysis of Speech,\" Journal of the Acoust. Society ofAmer., 87: 1738-1752, April, 1990.", | |
| "links": null | |
| }, | |
| "BIBREF6": { | |
| "ref_id": "b6", | |
| "title": "SVM based speaker verification using a GMM supervector kernel and nap variability compensation", | |
| "authors": [ | |
| { | |
| "first": "W", | |
| "middle": [], | |
| "last": "Campbell", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "D", | |
| "middle": [], | |
| "last": "Sturim", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "D", | |
| "middle": [], | |
| "last": "Reynolds", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Solomonoff", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2006, | |
| "venue": "Proc. ICASSP", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "97--100", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "W. Campbell, D. Sturim, D. Reynolds, and A. Solomonoff, \"SVM based speaker verification using a GMM supervector kernel and nap variability compensation,\" in Proc. ICASSP, Toulouse, France, 2006, pp. 97-100.", | |
| "links": null | |
| }, | |
| "BIBREF7": { | |
| "ref_id": "b7", | |
| "title": "Support vector machines using GMM supervectors for speaker verification", | |
| "authors": [ | |
| { | |
| "first": "W", | |
| "middle": [], | |
| "last": "Campbell", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "D", | |
| "middle": [], | |
| "last": "Sturim", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "D", | |
| "middle": [], | |
| "last": "Reynolds", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2006, | |
| "venue": "IEEE Signal Process. Lett", | |
| "volume": "13", | |
| "issue": "5", | |
| "pages": "308--311", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "W. Campbell, D. Sturim, and D. Reynolds, \"Support vector machines using GMM supervectors for speaker verification,\" IEEE Signal Process. Lett., vol. 13, no. 5, pp. 308-311, May 2006.", | |
| "links": null | |
| }, | |
| "BIBREF8": { | |
| "ref_id": "b8", | |
| "title": "An overview of text-independent speaker recognition: From features to supervectors", | |
| "authors": [ | |
| { | |
| "first": "T", | |
| "middle": [], | |
| "last": "Kinnunen", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "H", | |
| "middle": [], | |
| "last": "Li", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2010, | |
| "venue": "Speech Commun", | |
| "volume": "52", | |
| "issue": "1", | |
| "pages": "12--40", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "T. Kinnunen and H. Li, \"An overview of text-independent speaker recognition: From features to supervectors,\" Speech Commun., vol. 52, no. 1, pp. 12-40, 2010.", | |
| "links": null | |
| }, | |
| "BIBREF9": { | |
| "ref_id": "b9", | |
| "title": "State-of-the-art performance in text-independent speaker verification through open-source software", | |
| "authors": [ | |
| { | |
| "first": "B", | |
| "middle": [ | |
| "G B" | |
| ], | |
| "last": "Fauve", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "D", | |
| "middle": [], | |
| "last": "Matrouf", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "N", | |
| "middle": [], | |
| "last": "Scheffer", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J.-F", | |
| "middle": [], | |
| "last": "Bonastre", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J", | |
| "middle": [ | |
| "S D" | |
| ], | |
| "last": "Mason", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2007, | |
| "venue": "IEEE Trans. Audio, Speech, Lang. Process", | |
| "volume": "15", | |
| "issue": "7", | |
| "pages": "1960--1968", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "B. G. B. Fauve, D. Matrouf, N. Scheffer, J.-F. Bonastre, and J. S. D. Mason, \"State-of-the-art performance in text-independent speaker verification through open-source software,\" IEEE Trans. Audio, Speech, Lang. Process., vol. 15, no. 7, pp. 1960-1968, 2007.", | |
| "links": null | |
| }, | |
| "BIBREF10": { | |
| "ref_id": "b10", | |
| "title": "Channel compensation for SVM speaker recognition", | |
| "authors": [ | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Solomonoff", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "W", | |
| "middle": [ | |
| "M" | |
| ], | |
| "last": "Campbell", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "C", | |
| "middle": [], | |
| "last": "Quillen", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2004, | |
| "venue": "Proc. Odyssey04", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "57--62", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "A. Solomonoff, W. M. Campbell, and C. Quillen, \"Channel compensation for SVM speaker recognition,\" in Proc. Odyssey04, 2004, pp. 57-62.", | |
| "links": null | |
| }, | |
| "BIBREF11": { | |
| "ref_id": "b11", | |
| "title": "Comparison of scoring methods used in speaker recognition with joint factor analysis", | |
| "authors": [ | |
| { | |
| "first": "O", | |
| "middle": [], | |
| "last": "Glembek", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "L", | |
| "middle": [], | |
| "last": "Burget", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "N", | |
| "middle": [], | |
| "last": "Brummer", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "P", | |
| "middle": [], | |
| "last": "Kenny", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2009, | |
| "venue": "Proc. IEEE Int. Conf. Acoust., Speech, Signal Process", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "4057--4060", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "O. Glembek, L. Burget, N. Brummer, and P. Kenny, \"Comparison of scoring methods used in speaker recognition with joint factor analysis,\" in Proc. IEEE Int. Conf. Acoust., Speech, Signal Process., Taipei, Taiwan, Apr. 2009, pp. 4057-4060.", | |
| "links": null | |
| }, | |
| "BIBREF12": { | |
| "ref_id": "b12", | |
| "title": "A study of interspeaker variability in speaker verification", | |
| "authors": [ | |
| { | |
| "first": "P", | |
| "middle": [], | |
| "last": "Kenny", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "P", | |
| "middle": [], | |
| "last": "Ouellet", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "N", | |
| "middle": [], | |
| "last": "Dehak", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "V", | |
| "middle": [], | |
| "last": "Gupta", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "P", | |
| "middle": [], | |
| "last": "Dumouchel", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2008, | |
| "venue": "IEEE Trans. Audio, Speech, Lang. Process", | |
| "volume": "16", | |
| "issue": "5", | |
| "pages": "980--988", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "P. Kenny, P. Ouellet, N. Dehak, V. Gupta, and P. Dumouchel, \"A study of interspeaker variability in speaker verification,\" IEEE Trans. Audio, Speech, Lang. Process., vol. 16, no. 5, pp. 980-988, Jul. 2008.", | |
| "links": null | |
| }, | |
| "BIBREF13": { | |
| "ref_id": "b13", | |
| "title": "i-vector based Speaker Recognition on Short Utterances", | |
| "authors": [ | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Kanagasundaram", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "R", | |
| "middle": [], | |
| "last": "Vogt", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "D", | |
| "middle": [], | |
| "last": "Dean", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "S", | |
| "middle": [], | |
| "last": "Sridharan", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Mason", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2011, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "A. Kanagasundaram, R. Vogt, D. Dean, S. Sridharan, and M. Mason, \"i-vector based Speaker Recognition on Short Utterances,\" in Interspeech, 2011.", | |
| "links": null | |
| }, | |
| "BIBREF14": { | |
| "ref_id": "b14", | |
| "title": "Full-covariance ubm and heavy-tailed plda in i-vector speaker verification", | |
| "authors": [ | |
| { | |
| "first": "P", | |
| "middle": [], | |
| "last": "Matejka", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "O", | |
| "middle": [], | |
| "last": "Glembek", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "F", | |
| "middle": [], | |
| "last": "Castaldo", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "O", | |
| "middle": [], | |
| "last": "Plchot", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "P", | |
| "middle": [], | |
| "last": "Kenny", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "L", | |
| "middle": [], | |
| "last": "Burget", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Cernocky", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2011, | |
| "venue": "Proc. ICASSP '11", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "4828--4831", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "P. Matejka, O. Glembek, F. Castaldo, O. Plchot, P. Kenny, L. Burget, and J. Cernocky, \"Full-covariance ubm and heavy-tailed plda in i-vector speaker verification,\" Proc. ICASSP '11, pp. 4828-4831, 2011.", | |
| "links": null | |
| }, | |
| "BIBREF15": { | |
| "ref_id": "b15", | |
| "title": "i-vector with sparse representation classification for speaker verification", | |
| "authors": [ | |
| { | |
| "first": "J", | |
| "middle": [ | |
| "M K" | |
| ], | |
| "last": "Kua", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Epps", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "E", | |
| "middle": [], | |
| "last": "Ambikairajah", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2013, | |
| "venue": "Speech Commun", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "J.M.K. Kua, J. Epps, E. Ambikairajah, \"i-vector with sparse representation classification for speaker verification, \" Speech Commun, 2013.", | |
| "links": null | |
| }, | |
| "BIBREF16": { | |
| "ref_id": "b16", | |
| "title": "Sparse Representation for Signal Classification", | |
| "authors": [ | |
| { | |
| "first": "K", | |
| "middle": [], | |
| "last": "Huang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "S", | |
| "middle": [], | |
| "last": "Aviyente", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2006, | |
| "venue": "Neural Information Processing Systems", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "K. Huang and S. Aviyente, \"Sparse Representation for Signal Classification,\" Neural Information Processing Systems, 2006.", | |
| "links": null | |
| }, | |
| "BIBREF17": { | |
| "ref_id": "b17", | |
| "title": "Speaker verification using sparse representation classification", | |
| "authors": [ | |
| { | |
| "first": "J", | |
| "middle": [ | |
| "M K" | |
| ], | |
| "last": "Kua", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "E", | |
| "middle": [], | |
| "last": "Ambikairajah", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Epps", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "R", | |
| "middle": [], | |
| "last": "Togneri", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2011, | |
| "venue": "Proc. ICASSP", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "4548--4551", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "J. M. K. Kua, E. Ambikairajah, J. Epps, and R. Togneri, \"Speaker verification using sparse representation classification,\" in Proc. ICASSP, May 2011, pp. 4548-4551.", | |
| "links": null | |
| }, | |
| "BIBREF18": { | |
| "ref_id": "b18", | |
| "title": "Exemplar-based Sparse Representation and Sparse Discrimination for Noise Robust Speaker Identification", | |
| "authors": [ | |
| { | |
| "first": "R", | |
| "middle": [], | |
| "last": "Saeidi", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Hurmalainen", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "T", | |
| "middle": [], | |
| "last": "Virtanen", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "D", | |
| "middle": [ | |
| "A" | |
| ], | |
| "last": "Van Leeuwen", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2012, | |
| "venue": "Odyssey speaker and language recognition workshop", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "R. Saeidi, A. Hurmalainen, T. Virtanen, and D. A. van Leeuwen, \"Exemplar-based Sparse Representation and Sparse Discrimination for Noise Robust Speaker Identification,\" in Odyssey speaker and language recognition workshop, Singapore, 2012.", | |
| "links": null | |
| }, | |
| "BIBREF19": { | |
| "ref_id": "b19", | |
| "title": "Factor analysis of acoustic features using a mixture of probabilistic principal component analyzers for robust speaker verification", | |
| "authors": [ | |
| { | |
| "first": "T", | |
| "middle": [], | |
| "last": "Hasan", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J", | |
| "middle": [ | |
| "H L" | |
| ], | |
| "last": "Hansen", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2012, | |
| "venue": "Proc. Odyssey", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "T. Hasan and J. H. L. Hansen, \"Factor analysis of acoustic features using a mixture of probabilistic principal component analyzers for robust speaker verification,\" in Proc. Odyssey, Singapore, Jun. 2012.", | |
| "links": null | |
| }, | |
| "BIBREF20": { | |
| "ref_id": "b20", | |
| "title": "Mixtures of probabilistic principal component analyzers", | |
| "authors": [ | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Tipping", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "C", | |
| "middle": [], | |
| "last": "Bishop", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1999, | |
| "venue": "Neural Computation", | |
| "volume": "11", | |
| "issue": "2", | |
| "pages": "443--482", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "M. Tipping and C. Bishop, \"Mixtures of probabilistic principal component analyzers,\" Neural Computation, vol. 11, no. 2, pp. 443-482, 1999.", | |
| "links": null | |
| }, | |
| "BIBREF21": { | |
| "ref_id": "b21", | |
| "title": "Perceptual Factor Analysis for Speech Enhancement", | |
| "authors": [ | |
| { | |
| "first": "M", | |
| "middle": [ | |
| "Y" | |
| ], | |
| "last": "Lee", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2004, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "M. Y. Lee, Perceptual Factor Analysis for Speech Enhancement, Master Thesis, Institute of Computer Science and Information Engineering, National Cheng Kung University (2004).", | |
| "links": null | |
| }, | |
| "BIBREF22": { | |
| "ref_id": "b22", | |
| "title": "Tests of significance in factor analysis", | |
| "authors": [ | |
| { | |
| "first": "M", | |
| "middle": [ | |
| "S" | |
| ], | |
| "last": "Bartlett", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1950, | |
| "venue": "British Journal of Psychology, Statistical Section", | |
| "volume": "3", | |
| "issue": "", | |
| "pages": "77--85", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "M.S. Bartlett, \"Tests of significance in factor analysis,\" British Journal of Psychology, Statistical Section 3, 77-85, 1950", | |
| "links": null | |
| }, | |
| "BIBREF23": { | |
| "ref_id": "b23", | |
| "title": "K-SVD: an algorithm for designing overcomplete dictionaries for Sparse Representation", | |
| "authors": [ | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Aharon", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Elad", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Bruckstein", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2006, | |
| "venue": "IEEE Trans. Signal Process", | |
| "volume": "54", | |
| "issue": "11", | |
| "pages": "4311--4322", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "M. Aharon, M. Elad, and A. Bruckstein, \"K-SVD: an algorithm for designing overcomplete dictionaries for Sparse Representation,\" IEEE Trans. Signal Process., vol. 54, no. 11, pp. 4311-4322, Nov. 2006.", | |
| "links": null | |
| }, | |
| "BIBREF24": { | |
| "ref_id": "b24", | |
| "title": "Feature selection in face recognition: A sparse representation perspective", | |
| "authors": [ | |
| { | |
| "first": "A", | |
| "middle": [ | |
| "Y" | |
| ], | |
| "last": "Yang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Wright", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Y", | |
| "middle": [], | |
| "last": "Ma", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "S", | |
| "middle": [ | |
| "S" | |
| ], | |
| "last": "Sastry", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2007, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "A. Y. Yang, J. Wright, Y. Ma, and S. S. Sastry, \"Feature selection in face recognition: A sparse representation perspective,\" EECS Dept., Univ. California, Berkeley, CA, Tech. Report UCB/EECS-2007-99, 2007.", | |
| "links": null | |
| }, | |
| "BIBREF25": { | |
| "ref_id": "b25", | |
| "title": "Robust face recognition via sparse representation", | |
| "authors": [ | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Wright", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "A", | |
| "middle": [ | |
| "Y" | |
| ], | |
| "last": "Yang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Ganesh", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "S", | |
| "middle": [ | |
| "S" | |
| ], | |
| "last": "Sastry", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Y", | |
| "middle": [], | |
| "last": "Ma", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2009, | |
| "venue": "IEEE Transaction Pattern Analysis and Machine Intelligence", | |
| "volume": "31", | |
| "issue": "2", | |
| "pages": "210--226", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "J. Wright, A. Y. Yang, A. Ganesh, S. S. Sastry, and Y. Ma, \"Robust face recognition via sparse representation,\" IEEE Transaction Pattern Analysis and Machine Intelligence, vol. 31, no. 2, pp. 210-226, 2009.", | |
| "links": null | |
| }, | |
| "BIBREF26": { | |
| "ref_id": "b26", | |
| "title": "Pattern Classification", | |
| "authors": [ | |
| { | |
| "first": "R", | |
| "middle": [], | |
| "last": "Duda", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "P", | |
| "middle": [], | |
| "last": "Hart", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "D", | |
| "middle": [], | |
| "last": "Stork", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2000, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "R. Duda, P. Hart, and D. Stork, Pattern Classification, 2nd edition New York: Wiley, 2000.", | |
| "links": null | |
| }, | |
| "BIBREF27": { | |
| "ref_id": "b27", | |
| "title": "Face recognition based on nearest linear combinations", | |
| "authors": [ | |
| { | |
| "first": "S", | |
| "middle": [ | |
| "Z" | |
| ], | |
| "last": "Li", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1998, | |
| "venue": "IEEE Computer Society Conference on Computer Vision and Pattern Recognition", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "839--844", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "S. Z. Li, \"Face recognition based on nearest linear combinations,\" IEEE Computer Society Conference on Computer Vision and Pattern Recognition, pp. 839-844, 1998.", | |
| "links": null | |
| }, | |
| "BIBREF28": { | |
| "ref_id": "b28", | |
| "title": "Acquiring linear subspaces for face recognition under variable lighting", | |
| "authors": [ | |
| { | |
| "first": "K", | |
| "middle": [ | |
| "C" | |
| ], | |
| "last": "Lee", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Ho", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "D", | |
| "middle": [], | |
| "last": "Kriegman", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2005, | |
| "venue": "NIST 2005 Speaker Recognition Evaluation plan", | |
| "volume": "27", | |
| "issue": "", | |
| "pages": "684--698", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "K. C. Lee, J. Ho, and D. Kriegman, \"Acquiring linear subspaces for face recognition under variable lighting,\" IEEE Transaction Pattern Analysis and Machine Intelligence, vol. 27, no. 5, pp. 684-698, 2005. [30] NIST 2005 Speaker Recognition Evaluation plan, http://www.nist.gov/speech/tests/spk/2005/sre-05_evalplan-v6.pdf.", | |
| "links": null | |
| }, | |
| "BIBREF29": { | |
| "ref_id": "b29", | |
| "title": "Acoustic factor analysis for robust speaker verification", | |
| "authors": [ | |
| { | |
| "first": "T", | |
| "middle": [], | |
| "last": "Hasan", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J", | |
| "middle": [ | |
| "H L" | |
| ], | |
| "last": "Hansen", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2013, | |
| "venue": "IEEE Trans. Audio, Speech, and Lang. Process", | |
| "volume": "21", | |
| "issue": "4", | |
| "pages": "842--853", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "T. Hasan and J. H. L. Hansen, \"Acoustic factor analysis for robust speaker verification,\" IEEE Trans. Audio, Speech, and Lang. Process., vol. 21, no.4, pp.842-853, Apr. 2013.", | |
| "links": null | |
| }, | |
| "BIBREF30": { | |
| "ref_id": "b30", | |
| "title": "Kernel sparse representation-based classifier", | |
| "authors": [ | |
| { | |
| "first": "L", | |
| "middle": [], | |
| "last": "Zhang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "W", | |
| "middle": [ | |
| "D" | |
| ], | |
| "last": "Zhou", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "P", | |
| "middle": [ | |
| "C" | |
| ], | |
| "last": "Chang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Liu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Z", | |
| "middle": [], | |
| "last": "Yan", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "T", | |
| "middle": [], | |
| "last": "Wang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "F", | |
| "middle": [ | |
| "Z" | |
| ], | |
| "last": "Li", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2012, | |
| "venue": "IEEE Trans. Signal Processing", | |
| "volume": "60", | |
| "issue": "4", | |
| "pages": "1684--1695", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "L. Zhang, W. D. Zhou, P. C. Chang, J. Liu, Z. Yan, T. Wang, and F. Z. Li, \"Kernel sparse representation-based classifier,\" IEEE Trans. Signal Processing, vol. 60, no. 4, pp. 1684-1695, Apr. 2012.", | |
| "links": null | |
| }, | |
| "BIBREF31": { | |
| "ref_id": "b31", | |
| "title": "Weighted LDA techniques for i-vector based speaker verification", | |
| "authors": [ | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Kanagasundaram", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "D", | |
| "middle": [], | |
| "last": "Dean", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "R", | |
| "middle": [], | |
| "last": "Vogt", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Mclaren", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "S", | |
| "middle": [], | |
| "last": "Sridharan", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Mason", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2012, | |
| "venue": "IEEE International Conference on Acoustics, Speech and Signal Processing", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "4781--4784", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "A. Kanagasundaram, D. Dean, R. Vogt, M. McLaren, S. Sridharan, M. Mason, \"Weighted LDA techniques for i-vector based speaker verification,\" IEEE International Conference on Acoustics, Speech and Signal Processing, pp. 4781-4784, 2012.", | |
| "links": null | |
| }, | |
| "BIBREF32": { | |
| "ref_id": "b32", | |
| "title": "Robust Principal Component Analysis: Exact Recovery of Corrupted Low-Rank Matrices via Convex Optimization", | |
| "authors": [ | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Wright", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Ganesh", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "S", | |
| "middle": [], | |
| "last": "Rao", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Y", | |
| "middle": [], | |
| "last": "Ma", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2009, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "J. Wright, A. Ganesh, S. Rao, and Y. Ma, \"Robust Principal Component Analysis: Exact Recovery of Corrupted Low-Rank Matrices via Convex Optimization,\" Submitted to the Journal of the ACM, 2009.", | |
| "links": null | |
| } | |
| }, | |
| "ref_entries": { | |
| "TABREF0": { | |
| "content": "<table><tr><td>\u6280\u8853\u88ab\u7528\u4f86\u88dc\u511f\u4e0a\u8ff0\u63d0\u5230\u7684\u5dee\u7570\uff0c\u900f\u904e\u5c0d\u8a9e\u97f3\u53c3\u6578\u7684\u62c6\u89e3\uff0c\u5c07\u4e00\u4e9b\u8a9e\u8005\u4e0d\u76f8\u95dc\u7684\u8cc7\u8a0a\u522a</td></tr><tr><td>\u9664\uff0c\u5f97\u5230\u53bb\u9664\u5e72\u64fe\u5f8c\u6700\u80fd\u4ee3\u8868\u8a9e\u8005\u7684\u7279\u5fb5\u3002</td></tr><tr><td>\u8fd1\u5e74\u4f86\u5c0d\u65bc\u8907\u96dc\u7684\u8cc7\u8a0a(\u4f8b\u5982\u8a0a\u865f\u3001\u5716\u50cf)\uff0c\u5f80\u5f80\u5e0c\u671b\u53ef\u4ee5\u7528\u8f03\u7c21\u5316\u7684\u65b9\u5f0f\u5448\u73fe\uff0c\u7279</td></tr><tr><td>\u5225\u5728\u8a0a\u865f\u8655\u7406\u7684\u90e8\u5206\uff0c\u5206\u6790\u6642\u7d93\u5e38\u5148\u5c07\u8cc7\u6599\u8f49\u63db\u81f3\u4e0d\u540c\u7684\u5b9a\u7fa9\u57df\uff0c\u4e26\u4e14\u5047\u8a2d\u5176\u5728\u8f49\u63db</td></tr><tr><td>\u5f8c\uff0c\u6703\u5448\u73fe\u7a00\u758f\u5206\u5e03[24]\u3002\u5728\u8fd1\u671f\u7a00\u758f\u8868\u793a\u7684\u767c\u5c55\u4e2d\uff0cSRC \u5728[25][26]\u4e2d\u88ab\u63d0\u51fa\uff0cSRC</td></tr><tr><td>\u662f\u4e00\u500b Nonparametric \u5b78\u7fd2\u65b9\u6cd5\uff0c\u4e0d\u9700\u8a13\u7df4\u904e\u7a0b\u4f46\u662f\u9700\u8981\u8a13\u7df4\u8cc7\u6599\uff0c\u4ee5\u53ca\u53ef\u4ee5\u76f4\u63a5\u53c3\u8003</td></tr><tr><td>\u8a13\u7df4\u8cc7\u6599\u5c0d\u7b56\u662f\u8cc7\u6599\u9032\u884c\u5206\u985e\u7684\u52d5\u4f5c\u3002\u5be6\u9a57\u7d50\u679c\u986f\u793a\u51fa\u5728\u4eba\u81c9\u8fa8\u8b58\u7684\u61c9\u7528\u4e0a\uff0cSRC \u6709</td></tr><tr><td>\u8fd1\u5e7e\u5e74\u4e5f\u6709\u5c11\u6578\u7814\u7a76\u5c07 SRC \u61c9\u7528\u65bc\u8a9e\u8005\u9a57\u8b49[16-19]\u7684\u554f\u984c\u4e0a\uff0c\u7136\u800c\u6b64\u65b9\u9762\u7684\u7814\u7a76\u4ecd\u5c6c</td></tr><tr><td>\u65bc\u525b\u8d77\u6b65\u7684\u968e\u6bb5\uff0c\u4ecd\u6709\u8a31\u591a\u554f\u984c\u503c\u5f97\u6211\u5011\u63a2\u8a0e\u3002</td></tr><tr><td>\u672c\u8ad6\u6587\u63d0\u51fa\u4e00\u5957\u57fa\u65bc\u6838\u5316\u7a00\u758f\u8868\u793a\u7684\u8a9e\u8005\u8b58\u5225\u7cfb\u7d71\uff0c\u6587\u7ae0\u7684\u7de8\u6392\u5171\u5206\u6210\u4e03\u500b\u90e8\u5206:</td></tr></table>", | |
| "text": "\u793a\u4ee5\u53ca SVM \u5206\u985e\u65b9\u6cd5\u512a\u9ede\u7684\u8a9e\u8005\u8b58\u5225\u6f14\u7b97\u6cd5\u3002GMM \u8d85\u7d1a\u5411\u91cf\u662f\u4e00\u7a2e\u8a9e\u8005\u6a21\u578b\u7684\u5411\u91cf \u8868\u793a\u6cd5\uff0c\u901a\u7528\u80cc\u666f\u6a21\u578b(Universal Background Model, UBM)\u7d93\u904e\u8f38\u5165\u8a9e\u97f3\u8abf\u9069\u5f8c\u6210\u70ba GMM \u8a9e\u8005\u6a21\u578b\uff0c\u5176\u4e2d\u5404\u500b\u9ad8\u65af\u6210\u4efd(Components) \u7684\u671f\u671b\u503c\u88ab\u4e32\u5728\u4e00\u8d77\u5f62\u6210\u4e00\u500b\u8d85\u7d1a\u5411 \u91cf\u3002\u6b64\u5411\u91cf\u8868\u793a\u53ef\u4ee5\u7528\u4f86\u4ee3\u8868\u6574\u500b\u8a9e\u97f3\u6a94\u7684\u7279\u5fb5\u5411\u91cf\uff0c\u6700\u5f8c\uff0c\u518d\u4ee5 SVM \u4f5c\u70ba\u5206\u985e\u65b9\u6cd5 \u4f86\u9054\u5230\u8fa8\u8b58\u8a9e\u8005\u7684\u76ee\u7684\u3002\u7136\u800c\u8a9e\u8005\u8b58\u5225\u7cfb\u7d71\u6703\u56e0\u70ba\u9304\u88fd\u5de5\u5177\u548c\u80cc\u666f\u4e0d\u540c\uff0c\u9020\u6210\u901a\u9053\u5dee \u7570\u3001\u8aaa\u8a71\u5167\u5bb9\u5dee\u7570\u548c\u74b0\u5883\u5dee\u7570\u7684\u5e72\u64fe\uff0c\u5c0e\u81f4\u8fa8\u8b58\u7cfb\u7d71\u7684\u8fa8\u8b58\u7387\u4f4e\u843d\u3002\u64fe\u52d5\u5c6c\u6027\u6295\u5f71 (Nuisance Attribute Projection, NAP)\u662f\u4e00\u7a2e\u6709\u6548\u88dc\u511f\u901a\u9053\u5e72\u64fe\u7684\u65b9\u5f0f[7][11]\uff0c\u85c9\u7531\u8a2d\u8a08\u4e00 \u500b\u6700\u4f73\u5316\u554f\u984c\u4f7f\u5e72\u64fe\u6700\u5c0f\uff0c\u914d\u5408\u8457\u524d\u9762\u63d0\u5230\u7684\u65b9\u6cd5\uff0c\u66f4\u6709\u6548\u5730\u63d0\u9ad8\u8fa8\u8b58\u7cfb\u7d71\u7684\u5f37\u5065\u6027\u3002 \u5728 NAP \u4e4b\u5f8c\u9678\u7e8c\u6709\u806f\u5408\u56e0\u7d20\u5206\u6790(Joint Factor Analysis , JFA)[12][13]\u3001i-vector[14-16]\u7b49 \u8457\u6bd4 KNN (K-Nearest Neighbors)[27]\u4ee5\u53ca Nearest Subspace(NS)[28][29]\u66f4\u597d\u7684\u8fa8\u8b58\u7387\u3002", | |
| "html": null, | |
| "type_str": "table", | |
| "num": null | |
| }, | |
| "TABREF1": { | |
| "content": "<table><tr><td colspan=\"4\">Test)[22][23]\u7684\u6982\u5ff5\uff0c\u5efa\u7acb\u5047\u8aaa\uff0c\u627e\u5230\u81e8\u754c\u7684\u7279\u5fb5\u503c\u3002\u800c\u76ee\u524d\u8a9e\u8005\u8b58\u5225\u554f\u984c\u4e2d\uff0ci-vector</td></tr><tr><td colspan=\"4\">\u662f\u8868\u73fe\u6700\u597d\u7684\u53c3\u6578\u4e4b\u4e00\u3002\u85c9\u7531\u8a13\u7df4\u51fa\u7e3d\u9ad4\u8b8a\u7570\u77e9\u9663\uff0c\u5c07\u539f\u672c\u7684\u8d85\u7d1a\u5411\u91cf\u8f49\u5230\u66f4\u4f4e\u7dad\u7684\u7a7a</td></tr><tr><td colspan=\"4\">\u9593\uff0c\u4f7f i-vector \u66f4\u52a0\u8868\u73fe\u51fa\u8a9e\u8005\u53ca\u901a\u9053\u7684\u8cc7\u8a0a\uff0c\u56e0\u6b64\uff0c\u6211\u5011\u5c07 PPCA \u8d85\u7d1a\u5411\u91cf\u8f49\u63db\u5230\u7e3d</td></tr><tr><td colspan=\"4\">\u9ad4\u8b8a\u7570\u7a7a\u9593\u4e0a\uff0c\u5e0c\u671b\u5f97\u5230\u66f4\u5177\u9451\u5225\u529b\u7684 i-vector \u4f7f\u8fa8\u8b58\u6548\u80fd\u63d0\u5347\u3002</td></tr><tr><td colspan=\"4\">2.1 \u57fa\u65bc\u6a5f\u7387\u578b\u4e3b\u6210\u5206\u5206\u6790\u4e4b\u56e0\u7d20\u5206\u6790\u6a21\u578b</td></tr><tr><td colspan=\"4\">\u50b3\u7d71\u7684 GMM \u8d85\u7d1a\u5411\u91cf\uff0c\u4e26\u6c92\u6709\u8003\u91cf\u5230\u5176\u8072\u5b78\u7279\u5fb5\u53c3\u6578\u6709\u9ad8\u5ea6\u7684\u5197\u9918\u6027[21][31]\uff0c</td></tr><tr><td colspan=\"4\">\u56e0\u6b64\u61c9\u8a72\u63a1\u7528\u66f4\u4f4e\u7684\u5b50\u7a7a\u9593\u4f86\u8868\u793a\u3002\u5728[21][31]\u4e2d\u6bd4\u8f03\u6210\u4efd\u5206\u6790(Factor Analysis)\u8207\u4e3b\u6210</td></tr><tr><td colspan=\"4\">\u5206\u5206\u6790(PCA)\u7684\u95dc\u806f\u6027\uff0c\u89c0\u5bdf\u6210\u4efd\u5206\u6790\u7684\u6578\u5b78\u5f0f\uff1a</td></tr><tr><td>x \uf03d</td><td>Wz</td><td>\u03bc \uf02b</td><td>\u03b5 \uf02b</td></tr><tr><td/><td/><td/><td>Sparse</td></tr><tr><td colspan=\"4\">Representation Classifier, KSRC)\u3002\u7b2c\u4e94\u90e8\u5206\u70ba\u5be6\u9a57\u90e8\u5206\uff0c\u5c55\u793a\u63d0\u51fa\u4e4b\u6539\u826f\u65b9\u6cd5\u662f\u5426\u5177\u6709</td></tr><tr><td>\u5176\u5fc5\u8981\u6027\u3002\u6700\u5f8c\u5728\u7b2c\u516d\u90e8\u5206\u5247\u662f\u7d50\u8ad6\u3002</td><td/><td/><td/></tr><tr><td>\u4e8c\u3001\u53c3\u6578\u64f7\u53d6</td><td/><td/><td/></tr><tr><td colspan=\"4\">\u5728\u672c\u8ad6\u6587\u4e2d\uff0c\u6211\u5011\u5229\u7528\u6a5f\u7387\u578b\u4e3b\u6210\u5206\u5206\u6790\u5efa\u69cb\u8d85\u7d1a\u5411\u91cf[20][33]\uff0c\u4e26\u4ee5\u63d0\u51fa\u5df4\u96f7\u7279</td></tr><tr><td colspan=\"4\">\u6aa2\u5b9a(Bartlett Test)\u4e3b\u6210\u5206\u7684\u500b\u6578\u3002\u50b3\u7d71\u4e0a\uff0c\u8d85\u7d1a\u5411\u91cf\u662f\u7531\u9ad8\u65af\u6df7\u548c\u6a21\u578b\u5efa\u69cb\u800c\u6210\uff0c\u9019\u88e1</td></tr></table>", | |
| "text": "\u52a0\u5165\u4e3b\u6210\u5206\u5206\u6790\u7684\u6982\u5ff5\uff0c\u4e26\u5e0c\u671b\u80fd\u4ee5\u6a5f\u7387\u5206\u5e03\u6a21\u578b\u7684\u5f62\u5f0f\u8207\u9ad8\u65af\u6a21\u578b\u5c0d\u61c9\uff0c\u4f7f\u5f97\u8cc7\u6599\u9ede \u7531\u539f\u672c\u9ad8\u65af\u6df7\u548c\u6a21\u578b\u8f49\u6210 PPCA \u6df7\u548c\u6a21\u578b\uff0c\u518d\u900f\u904e Latent Factor \u7684\u8f49\u63db\uff0c\u5f62\u6210\u65b0\u7684\u8d85\u7d1a \u5411\u91cf\uff0c\u7a31\u4f5c PPCA \u8d85\u7d1a\u5411\u91cf[20][21][33]\uff0c\u5176\u4e2d\uff0c\u4e3b\u8ef8\u7684\u6311\u9078\uff0c\u6211\u5011\u5f15\u5165\u5df4\u96f7\u7279\u6aa2\u5b9a(Bartlett", | |
| "html": null, | |
| "type_str": "table", | |
| "num": null | |
| }, | |
| "TABREF2": { | |
| "content": "<table><tr><td colspan=\"6\">\u56db\u3001\u5b57\u5178\u8655\u7406\u53ca\u8b8a\u7570\u88dc\u511f \u5fb5\u53c3\u6578\uff0c\u4ee5\u56fa\u5b9a\u7dad\u5ea6\u7684\u5411\u91cf\u8868\u793a\u8a9e\u97f3\u8a0a\u865f\u3002\u5047\u8a2d\u6709 c \u500b\u4e0d\u540c\u8a9e\u8005\u985e\u5225\u7684\u8a13\u7df4\u8cc7\u6599\uff0c\u6bcf\u4e00</td></tr><tr><td colspan=\"6\">\u7b46\u8a13\u7df4\u8cc7\u6599\u70ba\u4e00\u500b i-vector\uff0c\u6211\u5011\u9700\u8981\u5efa\u69cb\u4e00\u500b\u8de8\u985e\u5225\u7684(Global)\u5b57\u5178 D\uff0c\u4f5c\u6cd5\u662f\u5c07\u6240\u6709</td></tr><tr><td colspan=\"6\">\u5728\u4ee5 i-vector \u70ba\u53c3\u6578\u7684\u67b6\u69cb\u4e0b\uff0ci-vector \u7684\u7e3d\u8b8a\u7570\u77e9\u9663\u662f\u5047\u8a2d\u8a13\u7df4\u8cc7\u6599\u6a19\u7c64\u4e0d\u540c\u6240\u8a13</td></tr><tr><td colspan=\"6\">\u7df4\u800c\u6210\uff0c\u56e0\u6b64\uff0c\u5b83\u5b58\u5728\u8207\u8a9e\u8005\u76f8\u95dc\u7684\u8b8a\u7570\uff0c\u9019\u662f\u6211\u5011\u60f3\u8981\u7684\uff0c\u540c\u6642\u5b83\u5b58\u5728\u4e00\u90e8\u5206\u9304\u97f3\u901a</td></tr><tr><td colspan=\"6\">\u9053\u53ca\u8aaa\u8a71\u5167\u5bb9\u9020\u6210\u7684\u8b8a\u7570\uff0c\u56e0\u6b64\uff0c\u4ee5\u5f80\u7684\u7814\u7a76\u90fd\u6703\u52a0\u5165 Linear Discriminant Analysis(LDA)</td></tr><tr><td colspan=\"6\">\u53ca Within-Class Covariance Normalization (WCCN)\u53bb\u88dc\u511f\u9304\u97f3\u901a\u9053\u7684\u8b8a\u7570[33]\uff0c\u5e0c\u671b\u5373\u4f7f</td></tr><tr><td colspan=\"6\">\u662f\u5728\u9304\u97f3\u901a\u9053\u4e0d\u540c\u7684\u72c0\u614b\u4e0b\uff0c\u627e\u51fa\u985e\u5225\u9593\u7684\u8b8a\u7570\u6700\u5927\u5316\uff0c\u985e\u5225\u5167\u7684\u8b8a\u7570\u6700\u5c0f\u5316\u7684\u7a7a\u9593\uff0c</td></tr><tr><td colspan=\"6\">Tw \uf02b m \uf03d \u4f7f\u8b8a\u7570\u6392\u9664\uff0c\u53e6\u5916\u8aaa\u8a71\u5167\u5bb9\u7684\u8b8a\u7570\u4e5f\u53ef\u4ee5\u7531 LDA \u7684\u8655\u7406\u89e3\u91cb\uff0c\u56e0\u70ba\u5b83\u5c07\u8a9e\u8005\u7684\u8aaa\u8a71\u5167 \u03bc (11)</td></tr><tr><td colspan=\"6\">\u5bb9\u505a\u8b8a\u7570\u6700\u5c0f\u7684\u5047\u8a2d\uff0c\u6d88\u9664\u8aaa\u8a71\u5167\u5bb9\u8b8a\u7570\u9020\u6210\u7684\u554f\u984c\u3002 m \u662f UBM \u8d85\u7d1a\u5411\u91cf\uff0c\u8207 JFA \u4e2d\u4f7f\u7528\u7684\u76f8\u540c\uff1bT \u4ee3\u8868\u7684\u662f\u6240\u6709\u8b8a\u7570\u6027\u7684\u77e9\u9663\uff1bi-vector \u672c\u7bc0\u63d0\u51fa\u5c0d i-vector \u5b57\u5178\u7684\u8655\u7406\u53ca\u88dc\u511f\u8fa6\u6cd5\uff0c\u5305\u62ec\u4ee5\u4f4e\u79e9\u77e9\u9663\u9084\u539f\u4ee5\u53ca\u6838\u5316\u7a00\u758f\u8868 \u4ee3\u8868\u7684\u662f\u7e3d\u9ad4\u8b8a\u7570\u6027\u5143\u7d20 w\u3002 \u6700\u5f8c\uff0c\u7e3d\u7d50\u53c3\u6578\u64f7\u53d6\u7684\u6574\u9ad4\u67b6\u69cb\uff0c\u53c3\u6578\u64f7\u53d6\u5305\u542b\u4e09\u500b\u90e8\u5206\uff0c\u7b2c\u4e00\uff0c\u7531 Universal \u793a\u5206\u985e\u5668(Kernel SRC)\u5206\u5225\u5c0d\u8aaa\u8a71\u5167\u5bb9\u53ca\u9304\u97f3\u901a\u9053\u8b8a\u7570\u505a\u88dc\u511f\uff0c\u4e26\u589e\u52a0\u5b57\u5178\u7684\u9451\u5225\u6027\u3002</td></tr><tr><td colspan=\"3\">4.1 \u4f4e\u79e9\u77e9\u9663\u9084\u539f\u5b57\u5178\u8655\u7406\u53ca\u8b8a\u7570\u88dc\u511f</td><td/><td/><td/></tr><tr><td>D</td><td>\uf03d</td><td>A</td><td>\uf02b</td><td>E</td><td>(15)</td></tr><tr><td colspan=\"4\">Sparse \u7684\u7279\u6027\uff0c\u56e0\u6b64\u6574\u7406\u6210\u4ee5\u4e0b\u7684\u6700\u4f73\u5316\u5f0f\uff1a</td><td/><td/></tr><tr><td colspan=\"6\">] \uf067 C D ) ...... ( A \uf02b rank 2 1 D D min [ , E A D \uf03d \u9032\u800c\uff0c\u6574\u7406\u6210\u4e00\u500b\u51f8\u6700\u4f73\u5316\u7684\u554f\u984c\uff0c\u4e26\u7528 Augmented Lagrange Multiplier (ALM)\u6c42\u89e3\u3002 (12) 0 E (16) \u5176\u4e2d\uff0cD ] ... [ 2 1 k x x x \uf03d x \u5982\u4e0b\u5f0f\uff1a \u900f\u904e\u4f4e\u79e9\u77e9\u9663\u9084\u539f\uff0c\u6211\u5011\u5c07\u6bcf\u500b\u8a9e\u8005\u985e\u5225\u7684\u5b57\u5178\u5206\u5225\u4f5c\u4f4e\u79e9\u77e9\u9663\u9084\u539f\u5206\u89e3\uff0c\u5f97\u5230\u8207 \u539f\u4f86\u5b57\u5178\u5927\u5c0f\u76f8\u540c\u7684\u4f4e\u79e9\u77e9\u9663 A\uff0c\u53d6\u4ee3\u539f\u4f86\u7684\u5b57\u5178\uff0c\u8b93\u5b57\u5178\u4e2d\u6bcf\u500b\u8a9e\u8005\u7684\u7279\u5fb5\u66f4\u70ba\u51f8\u986f\u3002 \uff0c\u4f7f\u539f\u8a0a\u865f\u8207\u91cd\u5efa\u8a0a\u865f\u4e4b\u9593\u7684\u8aa4\u5dee\u80fd\u8d8a\u5c0f\u8d8a\u597d\uff0c\u4e14 x \u8981\u7b26\u5408\u7a00\u758f\u7279\u6027\uff0c 1 2 2 min x Dx \uf06c \uf02b \uf02d x (13) 4.2 \u6838\u5316\u7a00\u758f\u8868\u793a\u5206\u985e\u5668 y \u5dee\uff0c\u7372\u5f97\u6700\u5c0f\u8aa4\u5dee\u8005\u5373\u70ba\u6240\u5c6c\u985e\u5225\uff1a \u85c9\u7531\u6838\u5316\u65b9\u6cd5(Kernel Trick)\uff0c\u7a00\u758f\u8868\u793a\u5206\u985e\u5668\u53ef\u9032\u4e00\u6b65\u975e\u7dda\u6027\u5316\uff0c\u7a31\u4f5c\u6838\u5316\u7a00\u758f\u8868 \u5f0f\u5206\u985e\u5668 [32]\u3002\u8f38\u5165\u7a7a\u9593\u7684\u8cc7\u6599\u7d93\u904e\u975e\u7dda\u6027\u6838\u5316\u6620\u5c04(Kernel Mapping)\u6295\u5c04\u81f3\u9ad8\u7dad\u53c3\u6578 \u7a7a\u9593\uff0c\u8b93\u539f\u672c\u5728\u8f38\u5165\u7a7a\u9593\u6df7\u6dc6\u7684\u53c3\u6578\u8cc7\u6599\u5728\u9ad8\u7dad\u7a7a\u9593\u8b8a\u6210\u53ef\u5206\u96e2\u7684\u3002\u900f\u904e\u66f4\u9032\u4e00\u6b65\u7684\u6838 \u5728\u89e3\u51fa\u7a00\u758f\u4fc2\u6578 x \u5f8c\uff0c\u6c7a\u7b56\u4e0a\uff0c\u5c07 k \u985e\u5225\u5404\u81ea\u7684\u5b57\u5178 D j \u53ca\u4fc2\u6578 x j \u9084\u539f y\uff0c\u4e26\u8207 y \u8a08\u7b97\u8aa4 2 2 * min arg j j j j x D \uf03d (14) \u5316\u964d\u7dad\u65b9\u6cd5\uff0c\u6211\u5011\u53ef\u5f97\u5230\u6e2c\u8a66\u8cc7\u6599\u7684\u7a00\u758f\u7d44\u5408\u4fc2\u6578(Sparse Combination Coefficients)\u4f86\u9032 \u884c\u5206\u985e\u7684\u52d5\u4f5c\u3002 y \uf02d \u6211\u5011\u5229\u7528\u7a00\u758f\u8868\u793a\u5177\u9451\u5225\u529b\u7684\u7279\u6027\uff0c\u61c9\u7528\u65bc\u8a9e\u8005\u8b58\u5225\u554f\u984c\u4e0a\uff0c\u5229\u7528 i-vector \u4f5c\u70ba\u7279</td></tr></table>", | |
| "text": "\u7684\u986f\u8457\u6c34\u5e73\u3002 2.3 i-vector \u555f\u767c\u65bc\u65e9\u671f JFA \u5728\u8a9e\u8005\u8b58\u5225\u4e0a\u7684\u61c9\u7528\uff0cDehak et al. \u63d0\u51fa\u7684\u4e00\u500b\u65b0\u7684\u5206\u6790\u65b9\u6cd5 i-vector [14-16]\uff0c\u4e0d\u50cf JFA \u5c07\u8a9e\u8005\u548c\u901a\u9053\u5206\u958b\uff0ci-vector \u50c5\u7528\u4e00\u500b\u7e3d\u9ad4\u8b8a\u7570\u6027\u7a7a\u9593\uff0c\u4ed6\u767c \u73fe JFA \u4e2d\u7684\u901a\u9053\u90e8\u5206\u4ecd\u5305\u542b\u4e86\u80fd\u7528\u4f86\u8b58\u5225\u8a9e\u8005\u7684\u8cc7\u8a0a\uff0c\u6240\u4ee5\u5c07 JFA \u4e2d\u5206\u958b\u7684\u8b8a\u7570\u90e8\u5206\uff0c \u5408\u4f75\u6210\u4e00\u500b\u55ae\u4e00\u7684\u7e3d\u9ad4\u8b8a\u7570\u6027\u7a7a\u9593\u8d85\u7d1a\u5411\u91cf\uff0c\u85c9\u7531\u5408\u4f75\u9304\u97f3\u65b9\u5f0f\u7684\u8b8a\u7570\u6027\uff0c\u63d0\u5347\u5176\u8fa8\u8b58 \u6027\uff0c\u8868\u793a\u5982\u4e0b\uff1a", | |
| "html": null, | |
| "type_str": "table", | |
| "num": null | |
| }, | |
| "TABREF3": { | |
| "content": "<table><tr><td colspan=\"6\">\u518d\u7d93\u7531\u8f49\u63db\u5230 i-vector \u4e0a\u3002\u85c9\u7531\u6bd4\u8f03\u5df4\u96f7\u7279\u6aa2\u5b9a\u8207\u6211\u5011\u5b9a\u7fa9\u7684\u4e3b\u8ef8\u9078\u53d6\u65b9\u5f0f\u4f86\u5224\u65b7\u5df4\u96f7 \u516d\u3001\u7d50\u8ad6</td></tr><tr><td colspan=\"6\">\u7279\u6aa2\u5b9a\u7684\u5fc5\u8981\u6027\u3002\u5728\u6211\u5011\u7684\u5be6\u9a57\u4e2d\uff0c\u5982\u8868\u4e00\u6240\u793a\uff0c\u56fa\u5b9a\u6bcf\u500b\u9ad8\u65af\u6210\u4efd\u7684\u4e3b\u8ef8\u500b\u6578\u70ba 27 \u9019 \u7bc7 \u8ad6 \u6587 \u63d0 \u51fa \u4e00 \u5957 \u57fa \u65bc \u7a00 \u758f \u8868 \u793a \u5206\u985e\u5668\u70ba \u57fa \u790e \u7684 \u8fa8 \u8b58 \u7cfb \u7d71 \uff0c \u5728 \u524d\u7aef\u4ee5 \u6642\u4f86\u5230\u6700\u9ad8 76.91%\uff0c\u4e0d\u904e\u96a8\u8457\u4e3b\u8ef8\u500b\u6578\u7684\u6e1b\u5c11\u8fa8\u8b58\u7387\u4e5f\u964d\u4f4e\u3002\u800c\u52a0\u5165\u5df4\u96f7\u7279\u6aa2\u5b9a\u5f8c\uff0c \u6211\u5011\u8a2d\u5b9a\u5df4\u96f7\u7279\u6aa2\u5b9a\u7684 05 . 0 \uf03d \uf061 PPCA-Supervector \u70ba\u53c3\u6578\uff0c\u52a0\u5165\u5df4\u96f7\u7279\u6aa2\u5b9a\u4f5c\u70ba\u6e96\u5247\uff0c\u6c7a\u5b9a\u6bcf\u500b\u9ad8\u65af Component \u6311\u9078\u4e3b \u3002\u5404\u500b\u9ad8\u65af\u6210\u4efd\u964d\u7dad\u7684\u6578\u76ee\u4e0d\u56fa\u5b9a\uff0c\u6709\u6548\u7684\u63d0\u5347\u8fa8\u8b58\u7387 \u8ef8\u7684\u8fa6\u6cd5\uff0c\u4f7f\u6bcf\u500b\u9ad8\u65af Component \u7684\u7dad\u5ea6\u53ef\u4ee5\u91dd\u5c0d\u8cc7\u6599\u7684\u4e0d\u540c\uff0c\u6c7a\u5b9a\u9069\u7576\u7684\u7dad\u5ea6\uff0c\u63a5 \u5230 77.01%\u3002\u5728\u63a5\u4e0b\u4f86\u7684\u5be6\u9a57\u4e2d\uff0c\u6211\u5011\u6703\u4ee5\u9019\u500b\u8fa8\u8b58\u7387\u70ba 77.01%\u7684\u7cfb\u7d71\u4f5c\u70ba\u5f8c\u7e8c\u5be6\u9a57\u7684 \u8457\uff0c\u8a13\u7df4\u51fa\u7e3d\u8b8a\u7570\u77e9\u9663\uff0c\u5c07 Supervector \u6295\u6620\u81f3\u7e3d\u8b8a\u7570\u7a7a\u9593\u4e0a\uff0c\u4ee5 i-vector \u4f5c\u70ba\u8fa8\u8b58\u53c3\u6578\u3002 \u57fa\u6e96(Baseline)\uff0c\u6e2c\u8a66\u63d0\u51fa/\u63a1\u7528\u7684\u6539\u9032\u65b9\u6cd5\u662f\u5426\u6709\u6548\u3002 \u5728\u5b57\u5178\u7684\u5efa\u69cb\u4e0a\uff0c\u6211\u5011\u63d0\u51fa\u4ee5\u4f4e\u79e9\u77e9\u9663\u9084\u539f\u4ee5\u53ca Kernel SRC \u9032\u884c\u8b8a\u7570\u88dc\u511f\uff0c\u53bb\u9664\u8aaa\u8a71</td></tr><tr><td colspan=\"6\">\u5167\u5bb9\u4ee5\u53ca\u901a\u9053\u8b8a\u7570\u9020\u6210\u7684\u5e72\u64fe\u3002\u5f9e\u5be6\u9a57\u7d50\u679c\u770b\u4f86\uff0cPPCA-Supervector \u5728\u672a\u505a\u5df4\u96f7\u7279\u6aa2\u5b9a</td></tr><tr><td colspan=\"6\">\u300a\u8868\u4e00\u300b\u4e0d\u540c\u4e3b\u8ef8\u500b\u6578\u8207\u5df4\u96f7\u7279\u6aa2\u5b9a\u9078\u53d6\u4e4b\u8fa8\u8b58\u7387\u6bd4\u8f03\u3002 \u524d\u5c31\u6709\u6bd4 GMM-Supervector \u597d\u7684\u6548\u679c\uff0c\u800c\u518d\u52a0\u5165\u5df4\u96f7\u7279\u6aa2\u5b9a\u5f8c\uff0c\u6548\u679c\u66f4\u52a0\u63d0\u9ad8\uff0c\u8207\u50b3</td></tr><tr><td colspan=\"6\">\u65b9\u6cd5 \u7d71\u57fa\u65bc i-vector \u7684\u8b58\u5225\u7cfb\u7d71\u76f8\u6bd4\uff0c\u8fa8\u8b58\u7387\u63d0\u5347\u4e86 10.07%\u3002\u6b64\u5916\uff0c\u518d\u52a0\u5165\u5169\u7a2e\u8b8a\u7570\u88dc\u511f\u65b9 \u8fa8\u8b58\u7387 76.60% 30\u500b\u4e3b\u8ef8 \u6cd5\u5f8c\uff0c\u4f4e\u79e9\u77e9\u9663\u9084\u539f\u4ee5\u53ca Kernel SRC \u5206\u5225\u53ef\u4ee5\u518d\u63d0\u5347\u7cfb\u7d71\u7684\u8fa8\u8b58\u7387 3.56%\u4ee5\u53ca 2.24%\u3002</td></tr><tr><td/><td/><td/><td>27\u500b\u4e3b\u8ef8</td><td/><td>76.91%</td></tr><tr><td>\u53c3\u8003\u6587\u737b</td><td colspan=\"2\">PPCA-SV</td><td>24\u500b\u4e3b\u8ef8</td><td>+ i-vector + SRC</td><td>76.40%</td></tr><tr><td/><td/><td/><td>21\u500b\u4e3b\u8ef8</td><td/><td>76.20%</td></tr><tr><td/><td/><td/><td>18\u500b\u4e3b\u8ef8</td><td/><td>75.99%</td></tr><tr><td/><td/><td/><td>15\u500b\u4e3b\u8ef8</td><td/><td>72.23%</td></tr><tr><td colspan=\"5\">PPCA-SV + Bartlett Test + i-vector + SRC</td><td>77.01%</td></tr><tr><td colspan=\"6\">\u5728\u4ee5\u4e0b\u5be6\u9a57\u4e2d\uff0c\u6211\u5011 \u6bd4\u8f03\u4e86\u56db\u7a2e\u4e0d\u540c\u8a9e\u8005\u8b58\u5225\u7cfb\u7d71\u3002 Baseline \u6211\u5011\u4f7f\u7528\u76ee\u524d</td></tr><tr><td colspan=\"6\">state-of-the-art \u8fa8\u8b58\u65b9\u6cd5\uff0ci-vector based cosine distance\uff0c\u5373\u4ee5 i-vector \u70ba\u53c3\u6578\u4e0b\uff0c\u5c07\u8f38\u5165</td></tr><tr><td colspan=\"6\">\u7279\u5fb5\u8207\u6bcf\u500b Class \u5b57\u5178\u7684 8 \u7b46 i-vector \u505a\u5167\u7a4d\u6c42\u5e73\u5747\uff0c\u6311\u9078\u5167\u7a4d\u6700\u5927\u7684\u8a9e\u8005\u985e\u5225\uff0c\u8868\u793a</td></tr><tr><td colspan=\"6\">\u76f8\u4f3c\u5ea6\u70ba\u6700\u5927\uff0c\u901a\u5e38\u5728\u53c3\u6578\u5f8c\u9762\u6703\u52a0\u4e0a LDA \u8207 WCCN \u5c0d\u9304\u97f3\u901a\u9053\u8b8a\u7570\u88dc\u511f\u3002\u5176\u9918\u4e09\u500b</td></tr><tr><td colspan=\"6\">\u7cfb\u7d71\u5206\u5225\u70ba\u57fa\u65bc\u7a00\u758f\u8868\u793a\u5668\u7684\u7cfb\u7d71\u3001\u57fa\u65bc\u6838\u5316\u7a00\u758f\u8868\u793a\u5668\u4f5c\u5b57\u5178\u88dc\u511f\u4e4b\u8b58\u5225\u7cfb\u7d71\u4ee5\u53ca\u57fa</td></tr><tr><td colspan=\"6\">\u65bc\u4f4e\u79e9\u77e9\u9663\u9084\u539f\u4e4b\u8b58\u5225\u7cfb\u7d71\u3002\u5be6\u9a57\u6578\u64da\u5982\u8868\u4e8c\u6240\u793a\uff0c\u900f\u904e\u4f4e\u79e9\u77e9\u9663\u9084\u539f\u65b9\u6cd5\u5efa\u69cb\u7684\u5b57\u5178</td></tr><tr><td colspan=\"6\">\u5177\u6709\u6700\u9ad8\u7684\u8fa8\u8b58\u7387 80.57%\uff0c\u6bd4 Baseline \u4ee5\u53ca\u6c92\u6709\u8b8a\u7570\u88dc\u511f\u7684\u7cfb\u7d71\u5206\u5225\u591a\u4e86 13.63%\u4ee5\u53ca</td></tr><tr><td colspan=\"6\">3.56%\u7684\u8fa8\u8b58\u7387\u3002\u8207\u6c92\u6709\u8b8a\u7570\u88dc\u511f\u7684\u7cfb\u7d71\u76f8\u6bd4\uff0cKernel LDA \u9054\u5230\u8b8a\u7570\u6027\u88dc\u511f\u7684\u6548\u679c\uff0c\u589e \u56e0\u6b64\uff0c\u6211\u5011\u900f\u904e Kernel LDA \u88dc\u511f\uff0c\u5e0c\u671b\u9304\u88fd\u65b9\u5f0f\u4e0d\u540c\u7522\u751f\u7684\u5e72\u64fe\u80fd\u85c9\u7531\u6700\u5927\u5316 Between-Class \u8b8a\u7570\u53ca\u6700\u5c0f\u5316 Within-Class \u8b8a\u7570\u5f97\u5230\u88dc\u511f\u3002 \u52a0\u4e86\u5b57\u5178\u7684\u9451\u5225\u6027\uff0c\u8fa8\u8b58\u7387\u63d0\u5347\u4e86 2.24%\u3002</td></tr><tr><td colspan=\"2\">\u4e94\u3001\u5be6\u9a57\u7d50\u679c</td><td colspan=\"4\">\u300a\u8868\u4e8c\u300b\u4e0d\u540c\u8a9e\u8005\u8b58\u5225\u7cfb\u7d71\u4e4b\u8fa8\u8b58\u7387\u6bd4\u8f03\u3002 \u65b9\u6cd5</td><td>\u8fa8\u8b58\u7387</td></tr><tr><td colspan=\"6\">\u6211\u5011\u4ee5 NIST2005 [30] \u505a\u70ba\u8a9e\u8005\u8cc7\u6599\u5eab\uff0cNIST \u6bcf\u5e74\u90fd\u6703\u9304\u88fd\u8a9e\u8005\u8cc7\u6599\u5eab\uff0c\u8a31\u591a\u7522 GMM-SV + i-vector + LDA + WCCN + CD (Baseline) 66.94%</td></tr><tr><td colspan=\"6\">\u696d\u754c\u3001\u5b78\u8853\u754c\u90fd\u6703\u4ee5\u6b64\u4f5c\u70ba\u8a55\u4f30\u6548\u80fd\u7684\u6a19\u6e96\uff0c\u800c 2005 \u5e74\u767c\u5e03\u7684\u8cc7\u6599\u5eab\u4ee5\u96fb\u8a71\u5c0d\u8a71\u8a9e\u97f3 PPCA-SV + Bartlett Test + i-vector + SRC 77.01%</td></tr><tr><td colspan=\"6\">\u6578\u64da\u70ba\u4e3b\uff0c\u540c\u6642\u6536\u96c6\u4e00\u4e9b\u8f14\u52a9\u9ea5\u514b\u98a8\u63a5\u6536\u7684\u6578\u64da\uff0c\u9019\u4e9b\u6578\u64da\u4e3b\u8981\u4f86\u81ea\u82f1\u8a9e\u6f14\u8b1b\uff0c\u4e26\u5305\u62ec PPCA-SV + Bartlett Test + i-vector + Kernel SRC 79.25%</td></tr><tr><td colspan=\"6\">\u56db\u7a2e\u984d\u5916\u8a9e\u8a00\u3002\u6211\u5011\u6311\u9078\u5176\u4e2d Male 8con-1con \u7684 Condition \u505a\u70ba\u6e2c\u8a66\u8cc7\u6599\u5eab\uff0c\u5176\u4e2d 8con PPCA-SV + Bartlett Test + i-vector + SRC (Low-Rank 80.57%</td></tr><tr><td colspan=\"4\">\u70ba\u8a13\u7df4\u8cc7\u6599\uff0c\u800c 1con \u662f\u6e2c\u8a66\u8cc7\u6599\u3002 Matrix Recovery Based Dictionary)</td><td/></tr><tr><td colspan=\"6\">\u70ba\u4e86\u9a57\u8b49\u5df4\u96f7\u7279\u6aa2\u5b9a\u662f\u5426\u80fd\u9032\u4e00\u6b65\u6539\u5584\u8fa8\u8b58\u7cfb\u7d71\uff0c\u6211\u5011\u5b9a\u7fa9\u53e6\u4e00\u5957\u4e3b\u8ef8\u500b\u6578\u9078\u53d6\u65b9</td></tr><tr><td colspan=\"6\">\u5f0f:\u5404\u500b\u9ad8\u65af\u6210\u4efd\u5728\u7d71\u4e00\u4e3b\u8ef8\u500b\u6578\u4e0b\u8207\u900f\u904e\u6aa2\u5b9a\u65b9\u5f0f\u6311\u9078\u9069\u5408\u4e3b\u8ef8\u500b\u6578\u4e0b\u7684\u5dee\u5225\uff0c\u6700\u5f8c</td></tr></table>", | |
| "text": "SRC(KSRC)\u5373\u7d50\u5408 Kernel LDA \u53ca SRC \u5169\u500b\u65b9\u6cd5\uff0c\u5c07\u539f\u59cb\u7279\u5fb5\u5411\u91cf\u7531 Kernel \u6295\u5f71\u81f3\u9ad8\u7dad\u5f8c\uff0c\u518d\u7531 LDA \u964d\u7dad\uff0c\u5728\u5148\u524d\u7684\u6587\u737b\u6709\u63d0\u5230\uff0ci-vector \u5b58\u5728 Channel \u7684\u8b8a\u7570\uff0c", | |
| "html": null, | |
| "type_str": "table", | |
| "num": null | |
| } | |
| } | |
| } | |
| } |