{ "paper_id": "O13-1023", "header": { "generated_with": "S2ORC 1.0.0", "date_generated": "2023-01-19T08:03:27.364998Z" }, "title": "", "authors": [], "year": "", "venue": null, "identifiers": {}, "abstract": "", "pdf_parse": { "paper_id": "O13-1023", "_pdf_hash": "", "abstract": [], "body_text": [ { "text": "\uf0e5 \uf03d \uf03d M i i i x b w x p 1 ) ( ) | ( \uf076 \uf076 \uf06c \u8a9e\u97f3\u9a57\u8b49\u6280\u8853\u4e3b\u8981\u7684\u4f5c\u6cd5\u6982\u8ff0\u5982\u4e0b\uff1a (a) \u95dc\u9375\u8a5e\u9a57\u8b49", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "", "sec_num": null }, { "text": "EQUATION", "cite_spans": [], "ref_spans": [], "eq_spans": [ { "start": 0, "end": 8, "text": "EQUATION", "ref_id": "EQREF", "raw_str": "\u5176\u4e2d M \u662f\u6df7\u5408\u6578\uff0c x \uf076 \u662f\u7dad\u5ea6\u70ba D \u7684\u7279\u5fb5\u5411\u91cf\uff0c ) (x b i \uf076 \u662f\u9ad8\u65af\u5206\u4f48\uff0c\u800c w i \u662f\u5404\u500b\u9ad8\u65af\u7684\u6b0a \u91cd\uff0c\u5fc5\u9808\u6eff\u8db3\u7684\u9650\u5236 \uf0e5 \uf03d \uf03d M i i w 1 1\u3002\u9ad8\u65af\u5206\u4f48\u7684\u5b9a\u7fa9\u5982\u4e0b\u6240\u793a\uff1a (2) \u5176\u4e2d i \uf06d \uf076 \u662f\u5e73\u5747\u5411\u91cf\uff0c i \uf053 \u662f\u5171\u8b8a\u7570\u77e9\u9663\u3002\u9ad8\u65af\u6df7\u5408\u6a21\u578b\u7684\u53c3\u6578\u53ef\u4ee5\u7528 \u03bb \u8868\u793a\u5982\u4e0b\uff1a \u03bb = {w i , i \uf06d \uf076 , i \uf053 } i = 1,\u2026,M", "eq_num": "(3)" } ], "section": "", "sec_num": null }, { "text": "\uf0fe \uf0fd \uf0fc \uf0ee \uf0ed \uf0ec \uf02d \uf053 \uf02d \uf02d \uf053 \uf03d \uf02d ) ( ) ( 2 1 exp | | ) 2 ( 1 ) ( 1 2 1 2 i i T i i D i x x x b \uf06d \uf06d \uf070 \uf076 \uf076 \uf076 \uf076 \uf076", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "", "sec_num": null }, { "text": "Proceedings of the Twenty-Fifth Conference on Computational Linguistics and Speech Processing (ROCLING 2013)", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "", "sec_num": null }, { "text": "EQUATION", "cite_spans": [], "ref_spans": [], "eq_spans": [ { "start": 0, "end": 8, "text": "EQUATION", "ref_id": "EQREF", "raw_str": "\uf0e5 \uf0e5 \uf03d \uf03d \uf03d T t t T t t t i x i p x x i p 1 1 ) , | ( ) , | ( \uf06c \uf06c \uf06d \uf076 \uf076 \uf076 \uf076 \uf05b \uf05d \uf05b \uf05d \u62d2\u7d55 \u63a5\u53d7 \uf071 \uf06c \uf06c \uf03c \uf0b3 \uf02d \uf03d ) | ( log ) | ( log ) | ( k k X p X p k X S \uf0fe \uf0fd \uf0fc \uf0ee \uf0ed \uf0ec \uf03d \uf0e5 \uf03d B b b k X p B X p 1 ) | ( 1 log ) | ( log \uf06c \uf06c \u81f3\u65bc\u8a9e\u8005\u6a21\u578b\u8a13\uf996\u7684\u904e\u7a0b\uff0c\u7c21\u55ae\u5730\u9673\u8ff0\u5982\u4e0b\uff1a a. \u521d\u59cb\u503c\u8a2d\u5b9a\uff1a\u5c07\u6bcf\u4e00\u4f4d\u8a9e\u8005\u5176\u8a13\u7df4\u8a9e\u6599\u4e4b\u7279\u5fb5\u5411\u91cf\uff0c\u6240\u8a08\u7b97\u51fa\u4f86\u7684\u5e73\u5747\u5411\uf97e\u53ca\u5171\u8b8a \uf962\u77e9\u9663\u4f5c\u70ba\u9ad8\u65af\u6df7\u5408\u6a21\u578b\u4e2d\u7b2c\u4e00\u500b\u9ad8\u65af\u5206\u4f48\u7684\uf96b\uf969\uff0c\u5176\u6df7\u5408\u6b0a\u91cd\u503c\u5247\u8a2d\u5b9a\u70ba 1.0\u3002 b. \u82e5\u6a21\u578b\u4e2d\u9ad8\u65af\u5206\u4f48\u7684\u7e3d\u6578\u5c0f\u65bc\u7cfb\u7d71\u6240\u8a2d\u5b9a\u7684\u6df7\u5408\u6578\uff0c\u5247\u9032\u884c\u9ad8\u65af\u5206\u4f48\u7684\u5206\u5272\u7a0b\u5e8f\u3002 \u672c\u7cfb\u7d71\u9078\u53d6\u6df7\u5408\u6b0a\u91cd\u503c\u6700\u9ad8\u7684\u9ad8\u65af\u5206\u4f48\u4f86\u9032\u884c\u5206\u5272\u3002 c. \u85c9\u7531\u671f\u671b\u503c\u6700\u5927\u5316(Expectation Maximization, EM) [13]\u6f14\u7b97\u6cd5\u53bb\u91cd\u65b0\u4f30\u6e2c\u6a21\u578b\uf96b\uf969 \u5305\u62ec\u6b0a\u91cd\u3001\u5e73\u5747\u5411\uf97e\u53ca\u5171\u8b8a\uf962\u77e9\u9663\u81f3\u7cfb\u7d71\u6240\u8a2d\u5b9a\u7684\u8fed\u4ee3\u6b21\u6578\u70ba\u6b62\u3002 d. \u91cd\u8907 b\u3001c \u6b65\u9a5f\uff0c\u76f4\u5230\u6a21\u578b\u4e2d\u9ad8\u65af\u5206\u4f48\u7684\u6df7\u5408\u6578\u5230\u9054\u7cfb\u7d71\u6240\u8a2d\u5b9a\u7684\u6df7\u5408\u6578\u70ba\u6b62\u3002 \u671f\u671b\u503c\u6700\u5927\u5316\u6f14\u7b97\u6cd5\u7684\u91cd\u4f30\u516c\u5f0f\u5982\u4e0b\uff1a (4) (5) (6) (7) \u6839\u64da\u8a13\u7df4\u597d\u7684\u9ad8\u65af\u6df7\u5408\u6a21\u578b \u03bb\uff0c\u6211\u5011\u5c0d\u65bc\u4efb\u4f55\u4e0d\u78ba\u5b9a\u7684\u8a9e\u97f3\u89c0\u6e2c\u5e8f\u5217 X \u53ef\u4ee5\u8a08\u7b97 P(X|\u03bb)\u3002 \u57fa\u65bc\u9ad8\u65af\u6df7\u5408\u6a21\u578b\u5c31\u53ef\u4ee5\u767c\u5c55\u51fa\u4e0d\u540c\u7684\u8a9e\u8005\u9a57\u8b49\u65b9\u6cd5\uff0c\u4f8b\u5982 Resenberg \u7b49\u4eba\u63d0\u51fa\u7fa4\u6b63\u898f \u5316\u8a08\u5206\u65b9\u6cd5[14-16]\uff0c\u4e8b\u5148\u7531\u975e\u5ba3\u544a\u8a9e\u8005\u7684\u8a9e\u6599\u6240\u8a13\u7df4\u51fa\u4f86\u7684\u4eff\u5192\u8005\u6a21\u578b\uff0c\u4ea6\u7a31\u70ba\u53cd\u8a9e\u8005 \u6a21\u578b(Anti-Speaker Model)\u3002\u6216\u662f\u5f9e\u5168\u90e8\u8a3b\u518a\u8a9e\u8005\u7684\u8a9e\u6599\u8a13\u7df4\u51fa\u4e00\u500b\u5171\u6709\u7684\u6a21\u578b\uff0c\u7a31\u70ba\u5168 \u57df\u8a9e\u8005\u6a21\u578b(Global Speaker Model)\u3002\u53cd\u8a9e\u8005\u6a21\u578b\u6216\u5168\u57df\u8a9e\u8005\u6a21\u578b\u53ef\u901a\u7a31\u70ba\u80cc\u666f\u8a9e\u8005\u6a21\u578b\uff0c \u800c\u6839\u64da\u6b63\u898f\u5316\u8a08\u5206\u6cd5\uff0c\u53ef\u4ee5\u4f7f\u7528\u5ba3\u544a\u8a9e\u8005\u6a21\u578b\u6a5f\u7387\u4e4b\u5c0d\u6578\u503c\u548c\u80cc\u666f\u8a9e\u8005\u6a21\u578b\u6a5f\u7387\u4e4b\u5c0d\u6578 \u503c\u76f8\u6e1b\uff0c\u4f5c\u70ba\u4e00\u6c7a\u7b56\u8b8a\u6578\u3002\u5982\u679c\u5176\u503c\u5927\u65bc\u4e00\u9580\u6abb\u503c \u03b8 \u5247\u63a5\u53d7\u70ba\u5ba3\u544a\u8a9e\u8005\u3002\u8a9e\u8005\u9a57\u8b49\u7684\u8a08 \u7b97\u516c\u5f0f\u5982\u4e0b\uff1a (8)", "eq_num": "(9)" } ], "section": "", "sec_num": null }, { "text": "EQUATION", "cite_spans": [], "ref_spans": [], "eq_spans": [ { "start": 0, "end": 8, "text": "EQUATION", "ref_id": "EQREF", "raw_str": "\u6211\u5011\u7a31\u516c\u5f0f(8)\u70ba\u7fa4\u6b63\u898f\u5316\u8a08\u5206\u51fd\u6578\uff0c\u5176\u4e2d k \uf06c \u4ee3\u8868\u5ba3\u544a\u8a9e\u8005 k \u7684\u53cd\u8a9e\u8005\u6a21\u578b\uff0cB \u5247\u70ba\u80cc \uf0e5 \uf03d \uf03d T t k t k x p T X p 1 ) | ( log 1 ) | ( log \uf06c \uf06c \uf076 \uf0e5 \uf03d \uf03d M k t k k t i i t x b w x b w x i p 1 ) ( ) ( ) , | ( \uf076 \uf076 \uf076 \uf06c \uf0e5 \uf03d \uf03d T t t i x i p T w 1 ) , | ( 1 \uf06c \uf076 2 1 1 2 2 ) , | ( ) , | ( i t T t T t t t i x i p x x i p \uf06d \uf06c \uf06c \uf073 \uf02d \uf03d \uf0e5 \uf0e5 \uf03d \uf03d \uf076 \uf076", "eq_num": "(10)" } ], "section": "", "sec_num": null }, { "text": "Proceedings of the Twenty-Fifth Conference on Computational Linguistics and Speech Processing (ROCLING 2013) ", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "", "sec_num": null }, { "text": "\uf028 \uf029 \uf028 \uf029 \uf028 \uf029 \uf028 \uf029 2 1 2 2 1 1 2 1 2 1 1 2 1 2 1 2 1 2 1 ln 2 1 2 8 1 , \uf053 \uf053 \uf053 \uf02b \uf053 \uf02b \uf02d \uf0f7 \uf0f8 \uf0f6 \uf0e7 \uf0e8 \uf0e6 \uf053 \uf02b \uf053 \uf02d \uf03d \uf02d t BA G G D \uf06d \uf06d \uf06d \uf06d \u666f\u8a9e\u8005\u4eba\u6578\u3002\u80cc\u666f\u8a9e\u8005\u6240\u63d0\u4f9b\u7684\u76f8\u4f3c\u5ea6\u6b63\u898f\u5316\u53ef\u4ee5\u62c9\u5927\u5ba3\u544a\u8a9e\u8005\u8207\u4eff\u5192\u8a9e\u8005\u4e4b\u76f8\u4f3c\u5ea6\u503c\uff0c \u4f7f\u5f97\u9580\u6abb\u503c\u80fd\u5920\u8f03\u5bb9\u6613\u88ab\u8a2d\u5b9a\u3002\u7531\u516c\u5f0f(8)\u8a08\u7b97\u6240\u5f97\u7684 S(X|k)\u503c\u518d\u8207\u9580\u6abb\u503c \u03b8 \u505a\u6bd4\u8f03\uff0c\u5224 \u65b7\u5176\u662f\u5426\u70ba\u5ba3\u544a\u8a9e\u8005\u3002 (\u4e00)\u53cd\u8a9e\u8005\u6a21\u578b\u9078\u64c7\u65b9\u6cd5 \u5728\u6b64\u6211\u5011\u662f\u4f9d\u64da\u8a9e\u8005\u6a21\u578b\u8ddd\u96e2\u6e2c\u91cf\u7684\u65b9\u6cd5\uff0c\u627e\u51fa\u8a9e\u8005\u5728\u8a9e\u8005\u6a21\u578b\u8cc7\u6599\u5eab\u4e2d\u7684\u540c\u8cea\u8a9e\u8005\u96c6 \u5408(Cohort Speaker Set)\u3002\u8ddd\u96e2\u6e2c\u91cf\u7684\u65b9\u6cd5\u5247\u662f\u63a1\u53d6 Bhattacharyya \u8ddd\u96e2\u4f86\u91cf\u6e2c\u8072\u5b78\u6a21\u578b\u4e4b \u9593\u7684\u8ddd\u96e2\u3002\u5047\u8a2d\u6211\u5011\u7d66\u5b9a\u5169\u500b\u9ad8\u65af\u5206\u4f48\uff0cG 1 =G(\u03bc 1 ;\u03a3 1 )\u53ca G 2 =G(\u03bc 2 ;\u03a3 2 )\uff0c\u5247\u5169\u500b\u9ad8\u65af\u5206\u4f48 \u4e4b\u9593\u7684 Bhattacharyya \u8ddd\u96e2\uff0c\u8a08\u7b97\u65b9\u5f0f\u5c31\u5982\u4e0b\u5f0f\u6240\u793a\uff1a (11) \u56e0\u70ba\u6bcf\u500b GMM \u6a21\u578b\u5305\u542b\u4e86\u591a\u500b\u9ad8\u65af\u6df7\u5408\uff0c\u6240\u4ee5\u8981\u8a08\u7b97\u5169\u500b\u8a9e\u8005 GMM \u6a21\u578b\u7684\u8ddd\u96e2\uff0c\u53ef \u4ee5\u5c0d\u5169\u7fa4 GMM \u6a21\u578b\u9593\u7684\u5169\u5169\u9ad8\u65af\u8ddd\u96e2(D BA (G 1 ,G 2 ))\u8a08\u7b97\u5176\u52a0\u6b0a\u548c\u3002GMM \u6a21\u578b\u53ef\u4ee5\u7528\u4f86 \u8861\u91cf\u5169\u8a9e\u8005\u8a9e\u97f3\u7684\u76f8\u4f3c\u5ea6\uff0c\u5728\u5f8c\u9762\u7684\u5be6\u9a57\u4e2d\u6211\u5011\u5c07\u4f7f\u7528\u5b83\u4f86\u6311\u9078\u548c\u5ba3\u544a\u8a9e\u8005\u8072\u97f3\u6700\u63a5\u8fd1 \u7684\u8a9e\u8005\uff0c\u4f5c\u70ba\u80cc\u666f\u8a9e\u8005\u3002 (\u4e8c)", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "", "sec_num": null }, { "text": "Proceedings of the Twenty-Fifth Conference on Computational Linguistics and Speech Processing(ROCLING 2013)", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "", "sec_num": null } ], "back_matter": [], "bib_entries": { "BIBREF0": { "ref_id": "b0", "title": "\u6b63\u4ea4\u5f0f\u9ad8\u65af\u6df7\u5408\u6a21\u578b\u4e4b\u8a9e\u8005\u9a57\u8b49\u7cfb\u7d71\uff02\uff0c\u4e2d\u8208\u5927\u5b78\u96fb\u6a5f\u5de5\u7a0b\u5b78\u7814\u7a76\u6240\u78a9 \u58eb\u8ad6\u6587\uff0c2009 \u5e74\u3002", "authors": [], "year": null, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "\u8b1d\u5fe0\u7a4e\uff0c\"\u6b63\u4ea4\u5f0f\u9ad8\u65af\u6df7\u5408\u6a21\u578b\u4e4b\u8a9e\u8005\u9a57\u8b49\u7cfb\u7d71\uff02\uff0c\u4e2d\u8208\u5927\u5b78\u96fb\u6a5f\u5de5\u7a0b\u5b78\u7814\u7a76\u6240\u78a9 \u58eb\u8ad6\u6587\uff0c2009 \u5e74\u3002", "links": null }, "BIBREF1": { "ref_id": "b1", "title": "\u9451\u5225\u5f0f\u8a13\uf996\u6cd5\u65bc\u8a9e\u8005\u9a57\u8b49\u4e4b\u7814\u7a76\uff02\uff0c\u4ea4\u901a\u5927\u5b78\u8cc7\u8a0a\u79d1\u5b78\u8207\u5de5\u7a0b\u7814\u7a76\u6240\u535a \u58eb\uf941\u6587\uff0c2009 \uf98e\u3002", "authors": [], "year": null, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "\u8d99\u6021\u7fd4\uff0c\"\u9451\u5225\u5f0f\u8a13\uf996\u6cd5\u65bc\u8a9e\u8005\u9a57\u8b49\u4e4b\u7814\u7a76\uff02\uff0c\u4ea4\u901a\u5927\u5b78\u8cc7\u8a0a\u79d1\u5b78\u8207\u5de5\u7a0b\u7814\u7a76\u6240\u535a \u58eb\uf941\u6587\uff0c2009 \uf98e\u3002", "links": null }, "BIBREF2": { "ref_id": "b2", "title": "\u542b\u8a9e\u8005\u9a57\u8b49\u4e4b\u5c0f\u578b\u5834\u6240\u4eba\u81c9\u8fa8\u8b58\u9580\u7981\u7cfb\u7d71\u4e4b\u7814\u767c\uff02\uff0c\u570b\u7acb\u6210\u529f\u5927\u5b78\u5de5\u7a0b \u79d1\u5b78\u7814\u7a76\u6240\u78a9\u58eb\u8ad6\u6587\uff0c2008 \u5e74\u3002", "authors": [], "year": null, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "\u8521\u4ef2\u9f61\uff0c\"\u542b\u8a9e\u8005\u9a57\u8b49\u4e4b\u5c0f\u578b\u5834\u6240\u4eba\u81c9\u8fa8\u8b58\u9580\u7981\u7cfb\u7d71\u4e4b\u7814\u767c\uff02\uff0c\u570b\u7acb\u6210\u529f\u5927\u5b78\u5de5\u7a0b \u79d1\u5b78\u7814\u7a76\u6240\u78a9\u58eb\u8ad6\u6587\uff0c2008 \u5e74\u3002", "links": null }, "BIBREF3": { "ref_id": "b3", "title": "An Overview of Automatic Speaker Recognition Technologies", "authors": [ { "first": "Reynolds", "middle": [], "last": "Douglas", "suffix": "" }, { "first": "A", "middle": [], "last": "", "suffix": "" } ], "year": 2002, "venue": "ICASSP", "volume": "", "issue": "", "pages": "4072--4075", "other_ids": {}, "num": null, "urls": [], "raw_text": "Reynolds Douglas A.,\"An Overview of Automatic Speaker Recognition Technologies\", ICASSP, p. 4072-4075, 2002.", "links": null }, "BIBREF4": { "ref_id": "b4", "title": "Speaker Recognition: A Tutorial\uff02", "authors": [ { "first": "Campbell", "middle": [], "last": "Joseph", "suffix": "" }, { "first": "P", "middle": [], "last": "", "suffix": "" } ], "year": 1997, "venue": "", "volume": "85", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Campbell Joseph P.,\"Speaker Recognition: A Tutorial\uff02, Vol. 85, No. 9, Proceedings of the IEEE, 1997.", "links": null }, "BIBREF5": { "ref_id": "b5", "title": "Spoken Language Processing", "authors": [ { "first": "X", "middle": [], "last": "Huang", "suffix": "" }, { "first": "A", "middle": [], "last": "Acero", "suffix": "" }, { "first": "H", "middle": [ "W" ], "last": "Hon", "suffix": "" } ], "year": 2001, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Huang, X., A. Acero, and H. W. Hon, Spoken Language Processing, Prentics Hall, New Jersey, 2001", "links": null }, "BIBREF6": { "ref_id": "b6", "title": "Generalized Mel Frequency Coefficients for Large-Vocabulary Speaker-Independent Continuous-Speech Recognition", "authors": [ { "first": "R", "middle": [], "last": "Vergin", "suffix": "" }, { "first": "D", "middle": [], "last": "O'shaughnessy", "suffix": "" }, { "first": "A", "middle": [], "last": "Farhat", "suffix": "" } ], "year": 1999, "venue": "IEEE Trans. on Speech and Audio Processing", "volume": "7", "issue": "5", "pages": "525--532", "other_ids": {}, "num": null, "urls": [], "raw_text": "R. Vergin and D. O'Shaughnessy and A. Farhat,\"Generalized Mel Frequency Coefficients for Large-Vocabulary Speaker-Independent Continuous-Speech Recognition\", IEEE Trans. on Speech and Audio Processing, Vol. 7, No. 5, pp. 525-532, September 1999.", "links": null }, "BIBREF7": { "ref_id": "b7", "title": "Confidence Measures for Speech Recognition: A Survey \uff02 , Speech Communication", "authors": [ { "first": "H", "middle": [], "last": "Jiang", "suffix": "" } ], "year": 2005, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Jiang H.,\"Confidence Measures for Speech Recognition: A Survey \uff02 , Speech Communication, 2005.", "links": null }, "BIBREF8": { "ref_id": "b8", "title": "Support Vector Machines Using GMM Supervectors for Speaker Verification", "authors": [ { "first": "W", "middle": [ "M" ], "last": "Campbell", "suffix": "" }, { "first": "", "middle": [], "last": "Etc", "suffix": "" } ], "year": 2006, "venue": "IEEE Signal Processing Letters", "volume": "13", "issue": "5", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Campbell W. M., etc,\"Support Vector Machines Using GMM Supervectors for Speaker Verification\", IEEE Signal Processing Letters, Vol. 13, No.5, 2006.", "links": null }, "BIBREF9": { "ref_id": "b9", "title": "\u61c9\u7528\u652f\u6490\u5411\u91cf\u6a5f\u65bc\u4e2d\u6587\u95dc\u9375\u8a5e\u9a57\u8b49\u4e4b\u7814\u7a76\uff02\uff0c\u570b\u7acb\u81fa\u7063\u79d1\u6280\u5927\u5b78\u8cc7\u8a0a\u7ba1 \u7406\u7814\u7a76\u6240\u78a9\u58eb\u8ad6\u6587\uff0c2007 \u5e74\u3002", "authors": [], "year": null, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "\u9ec3\u51a0\u9054\uff0c\"\u61c9\u7528\u652f\u6490\u5411\u91cf\u6a5f\u65bc\u4e2d\u6587\u95dc\u9375\u8a5e\u9a57\u8b49\u4e4b\u7814\u7a76\uff02\uff0c\u570b\u7acb\u81fa\u7063\u79d1\u6280\u5927\u5b78\u8cc7\u8a0a\u7ba1 \u7406\u7814\u7a76\u6240\u78a9\u58eb\u8ad6\u6587\uff0c2007 \u5e74\u3002", "links": null }, "BIBREF10": { "ref_id": "b10", "title": "Maximum Likelihood Linear Regression for Speaker Adaptation of Continuous Density Hidden Markov Models", "authors": [ { "first": "C", "middle": [ "J" ], "last": "Leggetter", "suffix": "" }, { "first": "P", "middle": [ "C" ], "last": "Woodland", "suffix": "" } ], "year": 1995, "venue": "Computer Speech and Language", "volume": "", "issue": "", "pages": "171--185", "other_ids": {}, "num": null, "urls": [], "raw_text": "Leggetter C. J. and Woodland P. C.,\"Maximum Likelihood Linear Regression for Speaker Adaptation of Continuous Density Hidden Markov Models\", Computer Speech and Language, p. 171-185, 1995.", "links": null }, "BIBREF11": { "ref_id": "b11", "title": "The Expectation-Maximization Algorithm", "authors": [ { "first": "T", "middle": [ "K" ], "last": "Moon", "suffix": "" } ], "year": 1996, "venue": "IEEE Signal Processing Magazine", "volume": "13", "issue": "6", "pages": "47--60", "other_ids": {}, "num": null, "urls": [], "raw_text": "T. K. Moon,\"The Expectation-Maximization Algorithm\", IEEE Signal Processing Magazine, Vol. 13, No. 6, pp. 47-60, November 1996.", "links": null }, "BIBREF12": { "ref_id": "b12", "title": "The Use of Cohort Normalized Scores for Speaker Recognition", "authors": [ { "first": "A", "middle": [ "E" ], "last": "Rosenberg", "suffix": "" }, { "first": "J", "middle": [], "last": "Delong", "suffix": "" }, { "first": "C", "middle": [ "H" ], "last": "Lee", "suffix": "" }, { "first": "B", "middle": [ "H" ], "last": "Juang", "suffix": "" }, { "first": "F", "middle": [ "K" ], "last": "Soong", "suffix": "" } ], "year": 1992, "venue": "", "volume": "", "issue": "", "pages": "599--602", "other_ids": {}, "num": null, "urls": [], "raw_text": "A. E. Rosenberg, J. Delong, C. H. Lee, B. H. Juang and F. K. Soong,\"The Use of Cohort Normalized Scores for Speaker Recognition\", Pro. ICSL 92. Banff, pp. 599-602. Oct. 1992.", "links": null }, "BIBREF13": { "ref_id": "b13", "title": "Speaker Verification Using Normalized Log-Likelihood Score", "authors": [ { "first": "Chi-Shi", "middle": [], "last": "Liu", "suffix": "" }, { "first": "Hsiao-Chuan", "middle": [], "last": "Wang", "suffix": "" }, { "first": "Chin-Hui", "middle": [], "last": "Lee", "suffix": "" } ], "year": 1996, "venue": "IEEE Trans. on Speech and Audio Processing", "volume": "", "issue": "", "pages": "57--60", "other_ids": {}, "num": null, "urls": [], "raw_text": "Chi-Shi Liu, Hsiao-Chuan Wang and Chin-Hui Lee,\"Speaker Verification Using Normalized Log-Likelihood Score\", IEEE Trans. on Speech and Audio Processing, pp. 57-60, Jan. 1996.", "links": null }, "BIBREF14": { "ref_id": "b14", "title": "\u6574\u5408\u9ad8\u65af\u6df7\u5408\u8207\u5177\u6027\u80fd\u6307\u6a19\u652f\u6490\u5411\uf97e\u6a5f\u6a21\u578b\u4e4b\u8a9e\u8005\u78ba\u8a8d\u7814\u7a76\uff02\uff0c\u570b\uf9f7\u4e2d \u592e\u5927\u5b78\u96fb\u6a5f\u5de5\u7a0b\u7814\u7a76\u6240\u78a9\u58eb\uf941\u6587\uff0c2008 \uf98e\u3002", "authors": [], "year": null, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "\u6e38\u667a\u7fd4\uff0c\"\u6574\u5408\u9ad8\u65af\u6df7\u5408\u8207\u5177\u6027\u80fd\u6307\u6a19\u652f\u6490\u5411\uf97e\u6a5f\u6a21\u578b\u4e4b\u8a9e\u8005\u78ba\u8a8d\u7814\u7a76\uff02\uff0c\u570b\uf9f7\u4e2d \u592e\u5927\u5b78\u96fb\u6a5f\u5de5\u7a0b\u7814\u7a76\u6240\u78a9\u58eb\uf941\u6587\uff0c2008 \uf98e\u3002", "links": null }, "BIBREF15": { "ref_id": "b15", "title": "Fisher Discriminant Analysis with Kernels", "authors": [ { "first": "S", "middle": [], "last": "Mika", "suffix": "" }, { "first": "G", "middle": [], "last": "R\u00e4tsch", "suffix": "" }, { "first": "J", "middle": [], "last": "Weston", "suffix": "" }, { "first": "B", "middle": [], "last": "Sch\u00f6lkopf", "suffix": "" }, { "first": "K", "middle": [ "R" ], "last": "M\u00fcller", "suffix": "" } ], "year": 1999, "venue": "Proc. Neural Networks for Signal Processing IX", "volume": "", "issue": "", "pages": "41--48", "other_ids": {}, "num": null, "urls": [], "raw_text": "Mika, S., G. R\u00e4tsch, J. Weston, B. Sch\u00f6lkopf, and K. R. M\u00fcller,\"Fisher Discriminant Analysis with Kernels\", in Proc. Neural Networks for Signal Processing IX, Madison, WI, USA, pp. 41-48, 1999.", "links": null } }, "ref_entries": { "TABREF2": { "num": null, "content": "
Proceedings of the Twenty-Fifth Conference on Computational Linguistics and Speech Processing (ROCLING 2013) Proceedings of the Twenty-Fifth Conference on Computational Linguistics and Speech Processing (ROCLING 2013) Proceedings of the Twenty-Fifth Conference on Computational Linguistics and Speech Processing (ROCLING 2013)
\u8868\u4e8c\u3001\u5c0d\u4e0d\u540c\u8fed\u4ee3\u6b21\u6578\u7684\u6548\u80fd\u8b8a\u5316 \u8868\u4e94\u3001\u5c0d\u4e0d\u540c\u6e2c\u8a66\u8a9e\u97f3\u8cc7\u6599\u6642\u9593\u9577\u5ea6\u7684\u6548\u80fd\u8b8a\u5316 \u73fe\u98fd\u548c\u72c0\u614b\u3002\u56e0\u6b64\u5728\u8003\u91cf\u517c\u9867\u8a9e\u8005\u9a57\u8b49\u7cfb\u7d71\u53cd\u61c9\u4e4b\u5373\u6642\u6027\uff0c\u4ee5\u53ca\u53ef\u4ee5\u900f\u904e\u7cfb\u7d71\u591a\u91cd\u9a57\u8b49 \u8fed\u4ee3\u6b21\u6578\u3001\u80cc\u666f\u8a9e\u8005\u4eba\u6578\u3001\u80cc\u666f\u8a9e\u8005\u6311\u9078\u65b9\u5f0f\u3001\u662f\u5426\u4f7f\u7528\u6027\u5225\u76f8\u95dc\u6a21\u578b\u3001\u6e2c\u8a66\u8a9e\u97f3\u8cc7\u6599
\u9a57\u8b49\u6548\u80fd\u8a08\u7b97 \u6839\u64da\u8a9e\u8005\u9a57\u8b49\u7684\u9a57\u8b49\u6d41\u7a0b\uff0c\u9a57\u8b49\u7684\u7d50\u679c\u53ef\u4ee5\u5206\u6210\u56db\u5927\u985e\uff0c\u5982\u5716\u4e09\u6240\u793a\u3002\u5176\u4e2d \u544a\u8a9e\u8005\u7684\u8a9e\u97f3\u88ab\u7cfb\u7d71\u62d2\u7d55\u7684\u500b\u6578(\u932f\u8aa4\u62d2\u7d55) \uff0cN 11 \u70ba\u5ba3\u544a\u8a9e\u8005\u7684\u8a9e\u97f3\u88ab\u7cfb\u7d71\u63a5\u53d7\u7684\u500b\u6578\uff0c N 10 \u4ee3\u8868\u5ba3 N 01 \u5247\u662f\u975e\u5ba3\u544a\u8a9e\u8005\u7684\u8a9e\u97f3\u88ab\u8aa4\u8a8d\u70ba\u5ba3\u544a\u8a9e\u8005\u800c\u63a5\u53d7\u7684\u500b\u6578(\u932f\u8aa4\u63a5\u53d7)\uff0cN 00 \u6307\u7684\u662f\u975e\u5ba3 \u544a\u8a9e\u8005\u7684\u8a9e\u97f3\u88ab\u7cfb\u7d71\u6b63\u78ba\u62d2\u7d55\u7684\u500b\u6578\u3002\u6839\u64da N 10 \u3001N 11 \u3001N 01 \u3001N 00 \u56db\u500b\u7d71\u8a08\u503c\uff0c\u6211\u5011\u53ef\u4ee5 \u5206\u5225\u8a08\u7b97\u7cbe\u78ba\u7387(Precision)\u3001\u53ec\u56de\u7387(Recall)\u4ee5\u53ca F \u5ea6\u91cf(F-Measure)\u5982\u4e0b\uff1a (12) (13) (14) \u5728\u8a9e\u8005\u9a57\u8b49\u7cfb\u7d71\u4e2d\uff0c\u9580\u6abb\u503c \u03b8 \u7684\u503c\u6703\u5f71\u97ff\u7cfb\u7d71\u7684\u6548\u80fd\u3002\u9580\u6abb\u503c\u8a2d\u5b9a\u7684\u904e\u9ad8\uff0c\u5bb9\u6613\u4f7f\u771f\u5be6 \u8a9e\u8005\u88ab\u7cfb\u7d71\u62d2\u7d55\uff0c\u4f7f\u5f97\u932f\u8aa4\u62d2\u7d55\u7387(False Rejection Rate)\u63d0\u9ad8\uff1b\u9580\u6abb\u503c\u8a2d\u5b9a\u7684\u904e\u4f4e\uff0c\u6703\u4f7f \u6b64\u5169\u96e3\u72c0\u6cc1\uff0c\u5728\u7cfb\u7d71\u8abf\u6821\u7684\u6642\u5019\u901a\u5e38\u6703\u4f7f\u7528\u6298\u8877\u7684\u6548\u80fd\u6307\u6a19 F \u503c\u9032\u884c\u6700\u4f73\u5316\uff0c\u6703\u627e\u5230 \u4f7f F \u503c\u6700\u5927\u7684\u9580\u6abb\u503c \u03b8 \u4f5c\u70ba\u6700\u7d42\u7cfb\u7d71\u8a2d\u5b9a\u7684\u9580\u6abb\u503c\u3002 \u6240\u793a\u3002\u7528\u6236\u7aef\u4f7f\u7528\u9ea5\u514b\u98a8\u9304\u88fd\u8a9e\u97f3\u4e26\u9032\u884c\u7aef\u9ede\u5075\u6e2c\uff0c\u4e26\u5c07\u9304\u88fd\u5230\u7684\u8a9e\u97f3\u4e32\u6d41\u900f\u904e Socket \u672c\u5be6\u9a57\u4e3b\u8981\u7684\u76ee\u7684\u662f\u627e\u51fa\u8a9e\u8005\u9a57\u8b49\u7cfb\u7d71\u6240\u9700\u8981\u4f7f\u7528\u4e4b\u9ad8\u65af\u6df7\u5408\u6a21\u578b\u5176\u6df7\u5408\u6578\u7684\u9069\u7576\u503c\u3002 \u7576\u6a21\u578b\u4e4b\u8fed\u4ee3\u6b21\u6578\u8a2d\u5b9a\u70ba 10 \u6642\uff0c\u5176\u5be6\u9a57\u7684\u7d50\u679c\u5982\u8868\u4e00\u6240\u793a\u3002\u6211\u5011\u53ef\u4ee5\u767c\u73fe\uff1a\u7576 GMM \u4e4b N \u4f4d\u8a9e\u8005\u6a21\u578b\u4f5c\u70ba\u5176\u80cc\u666f\u8a9e\u8005\u6a21\u578b\uff0cN \u5247\u70ba\u5176\u80cc\u666f\u8a9e\u8005\u4eba\u6578\u3002Random \u70ba\u5728\u6311\u9078\u80cc\u666f \u8a9e\u8005\u6642\u4e0d\u5206\u7537\u5973\u6027\u5225\uff0c\u800c\u4ee5\u96a8\u6a5f\u7684\u65b9\u5f0f\u6311\u9078 N \u4f4d\u8a9e\u8005\u6a21\u578b\u4f5c\u70ba\u5176\u80cc\u666f\u8a9e\u8005\u6a21\u578b\u3002 0.9427\uff1b\u800c\u7576\u8a13\u7df4\u8a9e\u97f3\u8cc7\u6599\u4e4b\u6642\u9593\u9577\u5ea6\u8d85\u904e 12000 \u500b\u97f3\u6846(\u7d04 120 sec)\u6642\uff0c\u7cfb\u7d71\u6548\u80fd\u5247\u5448 \u70ba\u4e86\u9054\u5230\u8f03\u4f73\u7684\u9a57\u8b49\u6548\u679c\uff0c\u6211\u5011\u4e5f\u4f5c\u4e86\u4e00\u7cfb\u5217\u7684\u5be6\u9a57\u4f86\u8abf\u6821\u7cfb\u7d71\u7684\u53c3\u6578\uff0c\u5305\u62ec\u6df7\u5408\u6578\u3001 0.8415\uff1b\u82e5\u8a13\u7df4\u8a9e\u97f3\u8cc7\u6599\u4e4b\u6642\u9593\u9577\u5ea6\u5230\u9054 6000 \u500b\u97f3\u6846(\u7d04 60 sec)\u6642\uff0c\u5247 F \u503c\u53ef\u4ee5\u5230\u9054 \u6838\u5fc3\u7684\u9a57\u8b49\u6a21\u7d44\u662f\u4ee5\u5206\u6563\u5f0f\u7684\u67b6\u69cb\u5206\u5225\u5be6\u73fe\u65bc Android \u7528\u6236\u7aef\u4ee5\u53ca\u9a57\u8b49\u4f3a\u670d\u5668\uff0c\u5982\u5716\u4e94 \u8a3b\uff1aB-Distance \u70ba\u5728\u6311\u9078\u80cc\u666f\u8a9e\u8005\u6642\u4e0d\u5206\u7537\u5973\u6027\u5225\uff0c\u800c\u4ee5\u8ddd\u96e2\u5ba3\u544a\u8a9e\u8005 GMM \u6a21\u578b\u6700\u8fd1 sec)\u4f86\u9032\u884c\u6bcf\u4e00\u4f4d\u8a9e\u8005\u4e4b\u9a57\u8b49\u6e2c\u8a66\u3002 \u4f86\u61c9\u7528\u7684\u91cd\u8981\u8da8\u52e2\u3002 \u5c31\u6703\u4e0b\u964d\u3002\u4f8b\u5982\u7576\u8a13\u7df4\u8a9e\u97f3\u8cc7\u6599\u4e4b\u6642\u9593\u9577\u5ea6\u70ba 1000 \u500b\u97f3\u6846(\u7d04 10 sec)\u6642\uff0c\u5176 F \u503c\u50c5\u6709 \u4eff\u5192\u8005\u5bb9\u6613\u88ab\u7cfb\u7d71\u8aa4\u5224\u70ba\u5ba3\u544a\u8a9e\u8005\uff0c\u4f7f\u5f97\u932f\u8aa4\u63a5\u53d7\u7387(False Acceptance Rate)\u4e0a\u5347\u3002\u7531\u65bc 01 11 11 N N N precision \uf02b \uf03d 10 11 11 N N N recall \uf02b \uf03d recall precision recall precision F \uf02b \u5716\u56db\u3001\u786c\u9ad4\u67b6\u69cb\u5716 \u7d04 15 sec)\u5247\u4f5c\u70ba\u6e2c\u8a66\u8a9e\u6599\u3002TCC-300 \u8a9e\u97f3\u8cc7\u6599\u5eab\u7684\u53d6\u6a23\u983b\u7387\u70ba 16kHz\uff0c\u8cc7\u6599\u578b\u614b\u70ba 16-bit Wav \u683c\u5f0f\u3002 (\u4e00)\u6df7\u5408\u6578\u7684\u5be6\u9a57 20 0.9398 0.8854 30 0.9556 0.8866 40 0.9582 0.9137 50 0.9716 0.9258 0.1 sec)\u6642\uff0c\u5176 F \u503c\u50c5\u6709 0.8407\uff1b\u82e5\u6e2c\u8a66\u8a9e\u97f3\u8cc7\u6599\u4e4b\u6642\u9593\u9577\u5ea6\u5230\u9054 100 \u500b\u97f3\u6846(\u7d04 1 sec) \u6642\uff0c\u5247 F \u503c\u53ef\u4ee5\u5230\u9054 0.9787\uff1b\u800c\u7576\u6e2c\u8a66\u8a9e\u97f3\u8cc7\u6599\u4e4b\u6642\u9593\u9577\u5ea6\u8d85\u904e 150 \u500b\u97f3\u6846(\u7d04 1.5 sec) \u6642\uff0c\u7cfb\u7d71\u6548\u80fd\u5247\u5448\u73fe\u98fd\u548c\u72c0\u614b\u3002\u56e0\u6b64\u5728\u8003\u91cf\u517c\u9867\u8a9e\u8005\u9a57\u8b49\u7cfb\u7d71\u53cd\u61c9\u4e4b\u5373\u6642\u6027\u53ca\u9a57\u8b49\u6548\u80fd \u7684\u60c5\u5f62\u4e0b\uff0c\u6211\u5011\u5c07\u672c\u5c08\u6848\u5be6\u969b\u7cfb\u7d71\u7684\u6e2c\u8a66\u8a9e\u97f3\u8cc7\u6599\u5176\u6642\u9593\u9577\u5ea6\u8a2d\u5b9a\u70ba 100 \u500b\u97f3\u6846(\u7d04 1 \u7d1a\u9700\u6c42\uff0c\u9a57\u8b49\u7cfb\u7d71\u53ef\u4ee5\u591a\u6b21\u8a62\u554f\u6216\u591a\u91cd\u9a57\u8b49\u7684\u65b9\u5f0f\u4f86\u63d0\u5347\u5b89\u5168\u6027\u3002\u6211\u5011\u76f8\u4fe1\u9019\u662f\u4e00\u500b\u672a \u4f46\u662f\u76f8\u5c0d\u5730\u8a08\u7b97\u91cf\u4ee5\u53ca\u6240\u9700\u6642\u9593\u4e5f\u6703\u589e\u52a0\uff1b\u8a13\u7df4\u8a9e\u97f3\u8cc7\u6599\u4e4b\u6642\u9593\u9577\u5ea6\u8f03\u77ed\u6642\uff0c\u7cfb\u7d71\u6548\u80fd \u5f48\u6027\u7684\u8a8d\u8b49\u65b9\u5f0f\u3002\u4f8b\u5982\u96f2\u7aef\u670d\u52d9\u4e2d\uff0c\u67e5\u8a62\u500b\u8cc7\u3001\u4fee\u6539\u5bc6\u78bc\u6216\u9032\u884c\u4ea4\u6613\u7b49\u6703\u6709\u4e0d\u540c\u5b89\u5168\u7b49 \u4e0a\u5347\u3002\u7531\u6b64\u6211\u5011\u53ef\u4ee5\u5f97\u77e5\uff1a\u8a13\u7df4\u8a9e\u97f3\u8cc7\u6599\u4e4b\u6642\u9593\u9577\u5ea6\u6108\u9577\u6108\u80fd\u63cf\u8ff0\u51fa\u8a9e\u8005\u7684\u767c\u8072\u7279\u6027\uff0c \u6027\u3002\u6b64\u5916\u6b64\u9a57\u8b49\u65b9\u5f0f\u5177\u6709\u5f48\u6027\uff0c\u672a\u4f86\u53ef\u9032\u4e00\u6b65\u7d50\u5408 RFID \u9a57\u8b49\u65b9\u5f0f\uff0c\u63d0\u4f9b\u96f2\u7aef\u670d\u52d9\u66f4\u6709 \u7d50\u679c\u5982\u8868\u516d\u6240\u793a\u3002\u6211\u5011\u53ef\u4ee5\u767c\u73fe\uff1a\u7576\u8a3b\u518a\u8a9e\u97f3\u8cc7\u6599\u4e4b\u6642\u9593\u9577\u5ea6\u6108\u9577\u6642\uff0c\u5176 F \u503c\u4e5f\u6703\u96a8\u4e4b \u6216\u5176\u8a9e\u97f3\u975e\u5ba3\u7a31\u8005\u672c\u4eba\u7684\u8a9e\u97f3\uff0c\u9a57\u8b49\u7cfb\u7d71\u5747\u53ef\u80fd\u52a0\u4ee5\u62d2\u7d55\uff0c\u56e0\u6b64\u53ef\u589e\u52a0\u9a57\u8b49\u7cfb\u7d71\u7684\u53ef\u9760 \u4e5f\u53ea\u6709\u4f7f\u7528\u4e00\u90e8\u4efd\u7684\u6e2c\u8a66\u8a9e\u97f3\u8cc7\u6599(100 \u500b\u97f3\u6846\uff0c\u7d04 1 sec)\u4f86\u9032\u884c\u9a57\u8b49\u6e2c\u8a66\u6642\uff0c\u5176\u5be6\u9a57\u7684 \u7d71\uff0c\u7cfb\u7d71\u53ef\u4ee5\u900f\u904e\u4f7f\u7528\u8005\u8f38\u5165\u7684\u5730\u5740\u8cc7\u6599\u4f86\u9032\u884c\u9a57\u8b49\uff0c\u5982\u679c\u4f7f\u7528\u8005\u4e0d\u77e5\u9053\u6b63\u78ba\u7684\u5730\u5740\u3001 \uf03d \u00d7 \u00d7 2 \u5716\u4e09\u3001Precision \u8207 Recall \u7684\u8a08\u7b97 \u4e09\u3001\u5206\u6563\u5f0f\u67b6\u69cb \u5728\u7db2\u969b\u7db2\u8def\u74b0\u5883\u4e0b\uff0c\u4f7f\u7528\u8005\u53ef\u80fd\u5f9e\u4efb\u610f\u5730\u9ede\u3001\u4efb\u610f\u88dd\u7f6e\u5b58\u53d6\u8cc7\u8a0a\uff0c\u56e0\u6b64\u9a57\u8b49\u7cfb\u7d71\u9808\u5c0d\u5404 \u7a2e\u884c\u52d5\u88dd\u7f6e\u63d0\u4f9b\u9a57\u8b49\u7684\u529f\u80fd\u3002\u7136\u800c\u5404\u7a2e\u884c\u52d5\u88dd\u7f6e\u7684\u8a08\u7b97\u80fd\u529b\u4e0d\u540c\uff0c\u8a9e\u97f3\u8fa8\u8b58\u53ca\u8a9e\u8005\u9a57\u8b49 \u7684\u6f14\u7b97\u6cd5\u9700\u8981\u9f90\u5927\u7684\u8a08\u7b97\u91cf\uff0c\u4e0d\u5bb9\u6613\u5728\u6240\u6709\u7684\u88dd\u7f6e\u4e0a\u505a\u5230\u5373\u6642\u6027\uff0c\u56e0\u6b64\u4e00\u500b\u5206\u6563\u5f0f\u7684\u7cfb \u7d71\u67b6\u69cb\u5177\u6709\u89e3\u6c7a\u6b64\u554f\u984c\u7684\u6f5b\u5728\u80fd\u529b\u3002\u6240\u4ee5\uff0c\u6211\u5011\u4ee5\u7528\u6236\u7aef\u53ca\u4f3a\u670d\u7aef\u7684\u67b6\u69cb\u4f86\u8a2d\u8a08\u4e00\u500b\u5206 \u6563\u5f0f\u7684\u8a9e\u97f3\u9a57\u8b49\u7cfb\u7d71\uff0c\u5176\u786c\u9ad4\u67b6\u69cb\u5982\u5716\u56db\u6240\u793a\u3002\u7528\u6236\u7aef\u70ba\u4f7f\u7528\u8005\u7684\u624b\u6301\u884c\u52d5\u88dd\u7f6e\u904b\u884c Android \u4f5c\u696d\u7cfb\u7d71\uff1b\u4f3a\u670d\u7aef\u5305\u542b\u4e86\u5169\u500b\u5be6\u9ad4\uff0c\u4e00\u500b\u662f\u9a57\u8b49\u4f3a\u670d\u5668\uff0c\u904b\u884c\u7684\u662f Windows 7 \u4f5c\u696d\u7cfb\u7d71\uff1b\u53e6\u4e00\u500b\u662f\u60a0\u904a\u96f2\u4f3a\u670d\u5668\uff0c\u63d0\u4f9b\u4f7f\u7528\u8005\u8cc7\u6599\u5eab(MySQL)\u7684\u5b58\u53d6\uff0c\u904b\u884c\u7684\u662f Linux Fedora \u4f5c\u696d\u7cfb\u7d71\u3002 Figure 1. The Dynamite \u5716\u4e94\u3001\u8a9e\u97f3\u9a57\u8b49\u6a21\u7d44\u7684\u5206\u6563\u5f0f\u67b6\u69cb\u5716 \u56db\u3001\u8a9e\u8005\u6a21\u578b\u5be6\u9a57 \u70ba\u4e86\u8a13\u7df4\u8f03\u4f73\u7684\u8a9e\u8005\u9a57\u8b49\u6a21\u578b\uff0c\u4ee5\u63d0\u5347\u5e73\u53f0\u9a57\u8b49\u6548\u80fd\uff0c\u6211\u5011\u4ee5 TCC-300 \u8a9e\u6599\u5eab\u9032\u884c\u5be6 \u9a57\u4f86\u8abf\u6821\u8a13\u7df4\u7684\u7a0b\u5e8f\u4ee5\u53ca\u53c3\u6578\u8a2d\u5b9a\u3002\u9996\u5148\uff0c\u6211\u5011\u91dd\u5c0d GMM \u6a21\u578b\u6240\u4f7f\u7528\u7684\u6df7\u5408\u6578\u3001\u4ee5\u53ca \u8a13\u7df4\u7684\u8fed\u4ee3\u6b21\u6578\u9032\u884c\u57fa\u790e\u5be6\u9a57\u3002\u5728\u4ee5\u4e0b\u5be6\u9a57\u4e2d\u82e5\u672a\u7279\u5225\u63d0\u53ca\uff0c\u5747\u662f\u4f7f\u7528\u6240\u6709\u975e\u5ba3\u544a\u8a9e\u8005 (\u5171 102 \u4f4d)\u7684\u6a21\u578b\u7d44\u6210\u53cd\u8a9e\u8005\u6a21\u578b\uff0c\u800c\u4f7f\u7528\u6a21\u578b\u6a5f\u7387\u7684\u7b97\u8853\u5e73\u5747\u4f5c\u70ba\u53cd\u6a21\u578b\u6a5f\u7387\u3002\u6211\u5011 \u5f9e TCC-300 \u8a9e\u97f3\u8cc7\u6599\u5eab\u4e2d\u9078\u53d6 103 \u4f4d\u8a9e\u8005\u6240\u9304\u88fd\u4e4b\u77ed\u53e5\u8a9e\u6599\u4f5c\u70ba\u5be6\u9a57\u6240\u4f7f\u7528\u7684\u8a13\u7df4\u53ca \u6e2c\u8a66\u8a9e\u6599\u3002\u5176\u4e2d\u7537\u6027\u8a9e\u8005\u6709 51 \u4f4d\uff0c\u5973\u6027\u8a9e\u8005\u6709 52 \u4f4d\uff0c\u5e73\u5747\u6bcf\u4f4d\u8a9e\u8005\u6709 65 \u53e5\u8a9e\u6599\u3002\u6bcf \u4f4d\u8a9e\u8005\u7684\u8a9e\u6599\u4e2d\u53d6 90%\u8a9e\u6599(\u7e3d\u9577\u5ea6\u5e73\u5747\u7d04 115 sec)\u4f86\u8a13\u7df4\u6a21\u578b\uff0c\u5176\u9918 10%(\u7e3d\u9577\u5ea6\u5e73\u5747 \u65af\u5206\u4f48\uff0c\u8a08\u7b97\u7684\u6642\u9593\u5c07\u6703\u5927\u5e45\u589e\u52a0\uff0c\u9ad8\u65af\u6df7\u5408\u6a21\u578b\u7684\u6578\u76ee\u4e5f\u5df2\u7d93\u8d85\u904e\u4e86\u8a9e\u97f3\u8fa8\u8b58\u6a21\u578b\u4e2d \u6240\u4f7f\u7528\u7684\u500b\u6578\u3002\u5728\u5206\u6563\u5f0f\u7684\u8a08\u7b97\u67b6\u69cb\u4e0b\uff0c\u9019\u6a23\u7684\u8907\u96dc\u5ea6\u96d6\u7136\u4ecd\u7136\u53ef\u4ee5\u505a\u5230\u5373\u6642\u8fa8\u8b58\uff0c\u4f46 \u662f\u70ba\u4e86\u6e1b\u5c11\u8a08\u7b97\u7684\u8ca0\u8f09\u91cf\uff0c\u6211\u5011\u9078\u64c7 Mixture 15 \u4f86\u9032\u884c\u5f8c\u7e8c\u7684\u5be6\u9a57\u3002\u56e0\u6b64\u5728\u8003\u91cf\u517c\u9867\u8a9e \u8005\u9a57\u8b49\u7cfb\u7d71\u53cd\u61c9\u4e4b\u5373\u6642\u6027\u53ca\u9a57\u8b49\u6548\u80fd\u7684\u60c5\u5f62\u4e0b\uff0c\u6211\u5011\u5c07\u672c\u5c08\u6848\u5be6\u969b\u7cfb\u7d71\u7684\u6a21\u578b\u6df7\u5408\u6578\u8a2d \u5b9a\u70ba 15 \u4f86\u8a13\u7df4\u6bcf\u4e00\u4f4d\u8a9e\u8005\u4e4b GMM \u6a21\u578b\u3002 \u8868\u4e00\u3001\u5c0d\u4e0d\u540c\u6df7\u5408\u6578\u7684\u6548\u80fd\u8b8a\u5316 Mixture Iteration Recall Precision F Measure 5 10 0.9104 0.9078 Mixture Iteration Recall Precision F Measure 15 5 0.9868 0.9796 0.9832 15 10 0.9883 0.9854 0.9868 15 15 0.9868 0.9868 0.9868 15 20 0.9868 0.9810 0.9839 15 25 0.9883 0.9854 0.9868 15 30 0.9883 0.9825 0.9854 15 35 0.9868 0.9912 0.9890 15 40 0.9883 0.9926 0.9904 15 45 0.9897 0.9868 0.9883 15 50 0.9897 0.9839 0.9868 (\u56db)\u4ee5\u6027\u5225\u4f86\u5340\u5206\u80cc\u666f\u8a9e\u8005\u4e4b\u5be6\u9a57 \u904e\u53bb\u5728\u8a9e\u97f3\u8fa8\u8b58\u4e0a\u4f7f\u7528\u548c\u6027\u5225\u76f8\u95dc\u7684\u8a9e\u97f3\u6a21\u578b\uff0c\u5c0d\u65bc\u8a9e\u97f3\u8fa8\u8b58\u7684\u6548\u80fd\u5747\u80fd\u5920\u7522\u751f\u63d0\u5347\u7684 Testing Data Frame Recall Precision F Measure 10 0.8253 0.8567 \u53ca\u8a3b\u518a\u8a9e\u97f3\u8cc7\u6599\u4e4b\u6642\u9593\u9577\u5ea6\u7b49\u3002\u5be6\u9a57\u7d50\u679c\u986f\u793a\u7576\u6df7\u5408\u6578\u70ba 15 \u4ee5\u4e0a\u3001\u8fed\u4ee3\u6b21\u6578\u70ba 10 \u4ee5\u4e0a \u6a5f\u5236\u4e2d\u5176\u4ed6\u4e0d\u540c\u7684\u9a57\u8b49\u65b9\u5f0f\u4f86\u4e92\u88dc\u5176\u6548\u80fd\u7684\u60c5\u5f62\u4e0b\uff0c\u6211\u5011\u5c07\u672c\u5c08\u6848\u5be6\u969b\u7cfb\u7d71\u7684\u8a3b\u518a\u8a9e\u97f3 \u6642\u5c31\u53ef\u4ee5\u9054\u5230\u7a69\u5b9a\u7684\u6548\u80fd\uff1b\u80cc\u666f\u8a9e\u8005\u4eba\u6578\u4e0a\u5347\u6642\u6548\u80fd\u53ef\u4ee5\u6301\u7e8c\u63d0\u5347\uff0c\u7576\u6e2c\u8a66\u8a9e\u97f3\u8cc7\u6599\u6216 \u8cc7\u6599\u5176\u6642\u9593\u9577\u5ea6\u8a2d\u5b9a\u70ba 6000 \u500b\u97f3\u6846(\u7d04 60 sec)\u4f86\u8a13\u7df4\u6bcf\u4e00\u4f4d\u8a9e\u8005\u4e4b GMM \u6a21\u578b\u3002 \u8a3b\u518a\u8a9e\u97f3\u8cc7\u6599\u4e4b\u6642\u9593\u9577\u5ea6\u589e\u52a0\u6642\uff0c\u4e5f\u53ef\u4ee5\u7522\u751f\u76f8\u540c\u7684\u6548\u679c\uff0c\u4f46\u662f\u8a08\u7b97\u91cf\u6703\u589e\u52a0\uff0c\u56e0\u6b64\u61c9 0.8407 \u8868\u516d\u3001\u5c0d\u4e0d\u540c\u8a13\u7df4\u8a9e\u97f3\u8cc7\u6599\u6642\u9593\u9577\u5ea6\u7684\u6548\u80fd\u8b8a\u5316 \u8003\u616e\u4f3a\u670d\u5668\u7684\u8a08\u7b97\u8ca0\u8f09\u662f\u5426\u904e\u91cd\u56e0\u800c\u5f71\u97ff\u8fa8\u8b58\u901f\u5ea6\uff1b\u80cc\u666f\u8a9e\u8005\u7684\u6311\u9078\u65b9\u5f0f\u5247\u986f\u793a\u4f7f\u7528 \u6548\u679c\u3002\u6211\u5011\u60f3\u63a2\u7a76\u5728\u8a9e\u8005\u8b58\u5225\u4e0a\u4f7f\u7528\u6027\u5225\u76f8\u95dc\u7684\u6a21\u578b\u5c0d\u65bc\u9a57\u8b49\u6548\u80fd\u662f\u5426\u4e5f\u80fd\u5920\u6709\u6240\u5e6b\u52a9\uff0c \u56e0\u6b64\u5c07\u7537\u5973\u7684\u8a9e\u6599\u5206\u958b\uff0c\u50c5\u5f9e\u8207\u6e2c\u8a66\u8a9e\u8005\u6027\u5225\u76f8\u540c\u7684\u8a9e\u8005\u4e2d\u6311\u9078\u80cc\u666f\u8a9e\u8005\u4f86\u9032\u884c\u5be6\u9a57\u3002 \u6211\u5011\u5c07\u6a21\u578b\u4e4b\u6df7\u5408\u6578\u8a2d\u5b9a\u70ba 15\u3001\u8fed\u4ee3\u6b21\u6578\u8a2d\u5b9a\u70ba 10\uff0c\u80cc\u666f\u8a9e\u8005\u7684\u6578\u76ee N \u5f9e 10 \u905e\u589e\u81f3 50\uff0c\u7be9\u9078\u80cc\u666f\u8a9e\u8005\u7684\u65b9\u5f0f\u5247\u662f\u63a1\u7528 Bhattacharyya \u8ddd\u96e2\u4f86\u7be9\u9078\u3002\u5728\u7be9\u9078\u80cc\u666f\u8a9e\u8005\u6642\uff0c\u6211 \u5011\u6703\u4ee5\u6e2c\u8a66\u8a9e\u8005\u7684\u6027\u5225(\u5047\u5b9a\u7cfb\u7d71\u5df2\u9810\u5148\u5224\u65b7\u6027\u5225)\u4f86\u9078\u53d6\u6027\u5225\u76f8\u540c\u7684\u80cc\u666f\u8a9e\u8005\uff0c\u6216\u5f9e\u5168 \u90e8\u8a9e\u8005\u4e2d\u7be9\u9078\u80cc\u666f\u8a9e\u8005\u4f86\u9032\u884c\u5be6\u9a57\uff0c\u5be6\u9a57\u7d50\u679c\u5982\u8868\u56db\u6240\u793a\u3002\u7531\u8868\u56db\u7d50\u679c\u53ef\u4ee5\u770b\u51fa\uff0c\u5728\u540c 20 0.9134 0.9311 0.9222 30 0.9339 0.9651 0.9493 40 0.9471 0.9743 0.9605 50 0.9633 0.9676 Bhattacharyya \u8ddd\u96e2\u6311\u9078\u548c\u5ba3\u7a31\u8a9e\u8005\u6700\u76f8\u8fd1\u7684\u8a9e\u8005\u4f5c\u70ba\u80cc\u666f\u8a9e\u8005\u9060\u8f03\u96a8\u6a5f\u65b9\u5f0f\u6311\u9078\u70ba\u4f73\uff1b Training Recall Precision F Measure \u6027\u5225\u76f8\u95dc\u7684\u5be6\u9a57\u5247\u986f\u793a\u53ea\u5f9e\u8207\u5ba3\u7a31\u8a9e\u8005\u540c\u6027\u5225\u7684\u8a9e\u8005\u4e2d\u6311\u9078\u80cc\u666f\u8a9e\u8005\uff0c\u6703\u6bd4\u5f9e\u6240\u6709\u7684\u8a9e Data Frame 1000 0.7915 0.8983 \u8005\u4e2d\u6311\u9078\u70ba\u4f73\uff0c\u4e5f\u5c31\u662f\u6027\u5225\u76f8\u95dc\u7684\u8a9e\u8005\u6a21\u578b\u5177\u6709\u8f03\u4f73\u7684\u9451\u5225\u529b\u3002\u6839\u64da\u5be6\u9a57\u7684\u7d50\u679c\uff0c\u6211\u5011 0.8415 \u4f7f\u7528\u8f03\u4f73\u7684\u8a13\u7df4\u6d41\u7a0b\u548c\u53c3\u6578\u8a2d\u5b9a\u4f86\u8a13\u7df4\u6a21\u578b\uff0c\u4e26\u61c9\u7528\u5728\u6211\u5011\u7684\u5c55\u793a\u7cfb\u7d71\u4e2d\u3002\u672a\u4f86\u5e0c\u671b\u5c07 0.9654 60 0.9486 0.9878 3000 0.9222 0.9235 0.9229 \u96f2\u7aef\u8a9e\u97f3\u9a57\u8b49\u5305\u88dd\u6210\u70ba\u7db2\u8def\u670d\u52d9\uff0c\u4ee5\u63d0\u4f9b\u7528\u6236\u5728\u7db2\u8def\u74b0\u5883\u4e2d\u65b9\u4fbf\u3001\u5feb\u901f\u3001\u5f48\u6027\u3001\u53ef\u9760\u5730 0.9678 6000 0.9178 0.9690 0.9427 \u8eab\u4efd\u8a8d\u8b49\u670d\u52d9\u3002 \u6a23\u80cc\u666f\u8a9e\u8005\u4eba\u6578\u4e0b\uff0c\u4ee5\u540c\u6027\u5225\u8a9e\u8005\u4f86\u9078\u53d6\u53cd\u8a9e\u8005\u6a21\u578b\u6703\u6bd4\u5f9e\u6240\u6709\u8a9e\u8005\u7be9\u9078\u70ba\u4f73\uff1b\u9019\u986f\u793a 70 0.9677 0.9720 0.9698 9000 0.9530 0.9701 0.9615 \u4e86\u540c\u6027\u5225\u7684\u80cc\u666f\u8a9e\u8005\u66f4\u80fd\u6b63\u78ba\u5730\u5340\u5225\u5ba3\u544a\u8a9e\u8005\u8207\u8fd1\u4f3c\u8a9e\u8005\uff0c\u4ea6\u5373\u6027\u5225\u76f8\u95dc\u7684\u6a21\u578b\u53ef\u4ee5\u6709 \u8f03\u4f73\u7684\u9451\u5225\u529b\u3002\u7136\u800c\u6211\u5011\u5fc5\u9808\u6ce8\u610f\u5230\uff0c\u6b64\u9a57\u8b49\u6548\u80fd\u7684\u63d0\u5347\u662f\u5728\u5047\u5b9a\u7cfb\u7d71\u5df2\u6b63\u78ba\u8fa8\u5225\u8a9e\u8005 \u6027\u5225\u7684\u689d\u4ef6\u4e0b\u800c\u9054\u6210\uff0c\u7cfb\u7d71\u7684\u8a2d\u8a08\u4e2d\u5fc5\u9808\u589e\u52a0\u53ef\u9760\u7684\u6027\u5225\u6c7a\u7b56\u6a21\u7d44\u3002 80 0.9780 0.9638 0.9708 90 0.9794 0.9709 0.9751 12000 0.9765 0.9779 0.9772 16000 0.9765 0.9794 \u81f4\u8b1d 0.9779 0.9091 10 10 0.9633 0.9719 0.9676 15 10 0.9883 0.9854 0.9868 20 10 0.9927 0.9927 0.9927 25 10 0.9927 0.9927 0.9927 30 10 0.9956 0.9941 0.9949 35 10 0.9956 0.9971 0.9963 40 10 0.9985 0.9985 0.9985 45 10 0.9985 0.9985 0.9985 50 10 0.9985 0.9985 0.9985 55 10 0.9985 0.9985 0.9985 60 10 1.0000 0.9985 (\u4e8c)\u8fed\u4ee3\u6b21\u6578\u7684\u5be6\u9a57 \u672c\u5be6\u9a57\u4e3b\u8981\u76ee\u7684\u662f\u627e\u51fa\u8a9e\u8005\u9a57\u8b49\u7cfb\u7d71\u6240\u9700\u8981\u4f7f\u7528\u4e4b\u9ad8\u65af\u6df7\u5408\u6a21\u578b\u5728\u6bcf\u6b21\u5206\u88c2\u5f8c\u4e4b\u8fed\u4ee3 \u6b21\u6578\u7684\u9069\u7576\u53c3\u6578\u503c\u3002\u7576\u6a21\u578b\u4e4b\u6df7\u5408\u6578\u8a2d\u5b9a\u70ba 15 \u6642\uff0c\u5176\u5be6\u9a57\u7d50\u679c\u5982\u8868\u4e8c\u6240\u793a\u3002\u7576\u8fed\u4ee3\u6b21 \u6578\u6108\u591a\u6642\uff0c\u5176 F \u503c\u4e26\u7121\u986f\u8457\u8b8a\u5316\uff0c\u5927\u7d04\u5728 0.98 \u81f3 0.99 \u4e4b\u9593\u3002\u53c8\u56e0\u6bcf\u6b21\u8fed\u4ee3\u5728\u8a13\u7df4\u6642\u90fd \u6703\u8017\u8cbb\u8f03\u591a\u6642\u9593\uff0c\u6240\u4ee5\u672c\u5c08\u6848\u5728\u8a13\u7df4\u6a21\u578b\u6642\uff0c\u5c07\u8fed\u4ee3\u6b21\u6578\u53c3\u6578\u56fa\u5b9a\u70ba 10\u3002 \u8fd1\u7684\u8a9e\u8005\uff0c\u800c\u8a13\u7df4\u51fa\u8f03\u5177\u6709\u9451\u5225\u529b\u7684\u6c7a\u7b56\u908a\u754c\u51fd\u6578\uff0c\u5176\u539f\u7406\u985e\u4f3c\u652f\u6490\u5411\u91cf\u6a5f\u5206\u985e\u5668\u3002\u8003 \u7121\u6cd5\u5c07\u6240\u6709\u8a9e\u8005\u6a21\u578b\u90fd\u7528\u4f86\u8a08\u7b97\u53cd\u8a9e\u8005\u6a21\u578b\u6a5f\u7387\u3002\u6b64\u6642\uff0c\u7be9\u9078\u80cc\u666f\u8a9e\u8005\u5c31\u662f\u8b93\u901f\u5ea6\u8207\u6548 \u9ad8\u65af\u6df7\u5408\u6a21\u578b\u8ddd\u96e2\u7684\u9069\u5207\u6027\u3002 \u8868\u4e09\u3001\u4ee5\u4e0d\u540c\u65b9\u5f0f\u4f86\u6311\u9078\u80cc\u666f\u8a9e\u8005\u6a21\u578b \u80cc\u666f\u8a9e\u8005\u4eba\u6578 B-Distance 10 0.9038 \u6642\u9593\u9577\u5ea6\u8f03\u77ed\u6642\uff0c\u7cfb\u7d71\u6548\u80fd\u5c31\u6703\u4e0b\u964d\u3002\u4f8b\u5982\u7576\u6e2c\u8a66\u8a9e\u97f3\u8cc7\u6599\u4e4b\u6642\u9593\u9577\u5ea6\u70ba 10 \u500b\u97f3\u6846(\u7d04 \u8fed\u4ee3\u6b21\u6578\u8a2d\u5b9a\u70ba 10\uff0c\u6bcf\u4e00\u4f4d\u8a9e\u8005\u53ea\u4f7f\u7528\u90e8\u4efd\u7684\u8a3b\u518a\u8a9e\u97f3\u8cc7\u6599\u4f86\u8a13\u7df4\u5176\u8a9e\u8005\u6a21\u578b\uff0c\u800c\u4e14 \u65b9\u6cd5\uff0c\u5728\u4e00\u500b\u5206\u6563\u5f0f\u7684\u7db2\u8def\u74b0\u5883\u4e2d\u9054\u5230\u4e86\u5373\u6642\u591a\u91cd\u9a57\u8b49\uff0c\u6211\u5011\u4e26\u88fd\u4f5c\u4e86\u4e00\u500b\u6280\u8853\u5c55\u793a\u7cfb 0.7310 \u80fd\u63cf\u8ff0\u51fa\u8a9e\u8005\u7684\u767c\u8072\u7279\u6027\uff0c\u4f46\u662f\u76f8\u5c0d\u5730\u8a08\u7b97\u91cf\u4ee5\u53ca\u6240\u9700\u6642\u9593\u4e5f\u6703\u589e\u52a0\uff1b\u6e2c\u8a66\u8a9e\u97f3\u8cc7\u6599\u4e4b \u7684\u6642\u9593\u9577\u5ea6\u5c0d\u65bc\u9a57\u8b49\u6548\u80fd\u6703\u7522\u751f\u4ec0\u9ebc\u6a23\u7684\u5f71\u97ff\uff0c\u56e0\u6b64\u6211\u5011\u5c07\u6a21\u578b\u4e4b\u6df7\u5408\u6578\u8a2d\u5b9a\u70ba 15\u3001 \u9032\u884c\u9a57\u8b49\u3002\u5169\u8005\u7684\u7d50\u5408\u53ef\u4ee5\u63d0\u9ad8\u8eab\u4efd\u9a57\u8b49\u7684\u53ef\u9760\u5ea6\u3002\u672c\u8a08\u756b\u6210\u529f\u5730\u7d50\u5408\u4e86\u4e0a\u8ff0\u5169\u7a2e\u9a57\u8b49 Random \u4f4d\u8a9e\u8005\u4f7f\u7528\u5176\u5168\u90e8\u7684\u8a3b\u518a\u8a9e\u97f3\u8cc7\u6599\u4f86\u8a13\u7df4\u5176\u8a9e\u8005\u6a21\u578b\uff0c\u800c\u53ea\u4f7f\u7528\u90e8\u4efd\u7684\u6e2c\u8a66\u8a9e\u97f3\u8cc7\u6599\u4f86 \u9032\u884c\u9a57\u8b49\u6e2c\u8a66\u6642\uff0c\u5176\u5be6\u9a57\u7684\u7d50\u679c\u5982\u8868\u4e94\u6240\u793a\u3002\u6211\u5011\u53ef\u4ee5\u767c\u73fe\uff1a\u7576\u6e2c\u8a66\u8a9e\u97f3\u8cc7\u6599\u4e4b\u6642\u9593\u9577 \u8a3b\u518a\u8a9e\u97f3\u8cc7\u6599\u7684\u6642\u9593\u9577\u5ea6\u6642\uff0c\u6703\u5c0d\u7cfb\u7d71\u6548\u80fd\u9020\u6210\u4ec0\u9ebc\u6a23\u7684\u5f71\u97ff\uff1f\u70ba\u4e86\u77ad\u89e3\u8a3b\u518a\u8a9e\u97f3\u8cc7\u6599 \u8a5e\u64f7\u53d6\u6280\u8853\u8207\u8a9e\u8005\u9a57\u8b49\u6280\u8853\u53ef\u5206\u5225\u4f7f\u7528 What one knows \u4ee5\u53ca Who one is \u7684\u9a57\u8b49\u65b9\u6cd5\u4f86 \u5ea6\u6108\u9577\u6642\uff0c\u5176 F \u503c\u4e5f\u6703\u96a8\u4e4b\u4e0a\u5347\u3002\u7531\u6b64\u6211\u5011\u53ef\u4ee5\u5f97\u77e5\uff1a\u6e2c\u8a66\u8a9e\u97f3\u8cc7\u6599\u4e4b\u6642\u9593\u9577\u5ea6\u6108\u9577\u6108 \u8853\u61c9\u7528\u5728\u8eab\u4efd\u8a8d\u8b49\u7cfb\u7d71\uff0c\u4ee5\u6539\u5584\u4f7f\u7528\u8005\u8a8d\u8b49\u670d\u52d9\u7684\u901f\u5ea6\u53ca\u6d41\u7a0b\uff0c\u8d8a\u4f86\u8d8a\u53d7\u5230\u91cd\u8996\u3002\u95dc\u9375 \u6642\uff0c\u7cfb\u7d71\u5373\u53ef\u5448\u73fe\u51fa\u4e0d\u932f\u7684\u9a57\u8b49\u6548\u80fd\u3002\u56e0\u800c\u4f7f\u6211\u5011\u60f3\u8981\u9032\u4e00\u6b65\u5730\u63a2\u8a0e\uff1a\u7576\u6211\u5011\u958b\u59cb\u6e1b\u5c11 \u5728\u8a9e\u97f3\u8fa8\u8b58\u548c\u9a57\u8b49\u6280\u8853\u9010\u6f38\u6210\u719f\u4ee5\u53ca\u500b\u4eba\u884c\u52d5\u88dd\u7f6e\u5feb\u901f\u666e\u53ca\u4e0b\uff0c\u5982\u4f55\u5c07\u7d50\u5408\u8a9e\u97f3\u76f8\u95dc\u6280 \u5f9e\u4e0a\u8ff0\u7684\u5be6\u9a57\u4e2d\u6211\u5011\u4e0d\u96e3\u767c\u73fe\uff1a\u7576\u6e2c\u8a66\u8a9e\u97f3\u8cc7\u6599\u4e4b\u6642\u9593\u9577\u5ea6\u8a2d\u5b9a\u70ba 100 \u500b\u97f3\u6846(\u7d04 1 sec) \u80fd\u9054\u5230\u6298\u8877\u7684\u53ef\u884c\u7b56\u7565\u3002\u672c\u5be6\u9a57\u4e5f\u540c\u6642\u9a57\u8b49\u4e86\u4ee5 Bhattacharyya \u8ddd\u96e2\u52a0\u6b0a\u65b9\u5f0f\u8a08\u7b97\u5169\u500b \u5168\u90e8\u7684\u6e2c\u8a66\u8a9e\u97f3\u8cc7\u6599\u4f86\u9032\u884c\u9a57\u8b49\u6e2c\u8a66\u3002\u70ba\u4e86\u77ad\u89e3\u6e2c\u8a66\u8a9e\u97f3\u8cc7\u6599\u7684\u6642\u9593\u9577\u5ea6\u5c0d\u65bc\u9a57\u8b49\u6548\u80fd \u6703\u7522\u751f\u4ec0\u9ebc\u6a23\u7684\u5f71\u97ff\uff0c\u56e0\u6b64\u6211\u5011\u5c07\u6a21\u578b\u4e4b\u6df7\u5408\u6578\u8a2d\u5b9a\u70ba 15\u3001\u8fed\u4ee3\u6b21\u6578\u8a2d\u5b9a\u70ba 10\uff0c\u6bcf\u4e00 \u4e94\u3001\u7d50\u8ad6 (\u516d)\u8a3b\u518a\u8a9e\u97f3\u8cc7\u6599(\u8a13\u7df4\u8cc7\u6599)\u6642\u9593\u9577\u5ea6\u4e4b\u5be6\u9a57 \u91cf\u7cfb\u7d71\u5be6\u969b\u904b\u4f5c\u6642\uff0c\u8a3b\u518a\u4eba\u6578\u53ef\u80fd\u6703\u96a8\u8457\u7cfb\u7d71\u4f7f\u7528\u6642\u9593\u905e\u589e\uff1b\u82e5\u8a9e\u8005\u7e3d\u6578\u5f88\u5927\u6642\uff0c\u52e2\u5fc5 \u65bc\u4e0a\u8ff0\u7684\u5404\u9805\u5be6\u9a57\u4e2d\uff0c\u6bcf\u4e00\u4f4d\u8a9e\u8005\u90fd\u4f7f\u7528\u4e86\u5176\u5168\u90e8\u7684\u8a3b\u518a\u8a9e\u97f3\u8cc7\u6599\u4f86\u8a13\u7df4\u5176\u8a9e\u8005\u6a21\u578b\uff0c 0.9993 (\u4e09)\u80cc\u666f\u8a9e\u8005\u4eba\u6578\u8207\u6311\u9078\u65b9\u5f0f\u4e4b\u5be6\u9a57 \u70ba\u4e86\u77ad\u89e3\u80cc\u666f\u8a9e\u8005\u6578\u76ee\u4ee5\u53ca\u7be9\u9078\u65b9\u5f0f\u5c0d\u65bc\u9a57\u8b49\u6548\u80fd\u7684\u5f71\u97ff\uff0c\u6211\u5011\u4f7f\u7528\u4e86\u5169\u7a2e\u6311\u9078\u80cc\u666f\u8a9e \u8005\u7684\u65b9\u5f0f\uff0c\u4e00\u7a2e\u662f\u6839\u64da Bhattacharyya \u8ddd\u96e2\u627e\u51fa\u548c\u5ba3\u544a\u8a9e\u8005\u6700\u76f8\u8fd1\u7684 N \u540d\u8a9e\u8005\u505a\u70ba\u53cd\u8a9e \u8005\u6a21\u578b\uff0c\u53e6\u4e00\u7a2e\u5247\u662f\u4ee5\u96a8\u6a5f\u65b9\u5f0f\u6311\u9078\u3002\u6211\u5011\u5c07\u6a21\u578b\u4e4b\u6df7\u5408\u6578\u8a2d\u5b9a\u70ba 15\u3001\u8fed\u4ee3\u6b21\u6578\u8a2d\u5b9a \u70ba 10\uff0c\u80cc\u666f\u8a9e\u8005\u7684\u6578\u76ee N \u5f9e 10 \u905e\u589e\u81f3 50 \u6642\uff0cF \u5ea6\u91cf\u7684\u5be6\u9a57\u7d50\u679c\u5982\u8868\u4e09\u6240\u793a\u3002\u7531\u8868\u4e09 \u53ef\u4ee5\u770b\u51fa\uff0c\u7576\u80cc\u666f\u8a9e\u8005\u4eba\u6578\u6108\u591a\u6642\uff0c\u5176 F \u503c\u4e5f\u6703\u96a8\u4e4b\u4e0a\u5347\u3002\u7531\u6b64\u6211\u5011\u53ef\u4ee5\u5f97\u77e5\uff1a\u80cc\u666f\u8a9e \u6a19\u8a18\u70ba B-Distance)\u9078\u53d6\u80cc\u666f\u8a9e\u8005\u7684\u65b9\u5f0f\u6703\u964d\u81f3 0.9038\uff0c\u4f46\u5f88\u660e\u986f\u5730\u4ecd\u7136\u6bd4\u4ee5\u96a8\u6a5f\u7684\u65b9\u5f0f \u4f86\u6311\u9078\u53cd\u8a9e\u8005\u7684\u6548\u80fd\u70ba\u4f73\u3002\u9019\u986f\u793a\u4e86 Bhattacharyya \u8ddd\u96e2\u80fd\u5920\u6b63\u78ba\u5730\u627e\u51fa\u548c\u5ba3\u544a\u8a9e\u8005\u76f8 \u7684\u6642\u9593\u3002 \u7684\u6642\u9593\u3002 (\u4e94)\u6e2c\u8a66\u8a9e\u97f3\u8cc7\u6599\u6642\u9593\u9577\u5ea6\u4e4b\u5be6\u9a57 (3) \u6839\u64da\u672c\u5c08\u6848\u7cfb\u7d71\u4e2d\u8a9e\u97f3\u8cc7\u6599\u53d6\u6a23\u6642\u6240\u63a1\u53d6\u7684\u97f3\u6846\u64f7\u53d6\u65b9\u5f0f\uff0c100 \u500b\u97f3\u6846\u7d04\u7b49\u65bc 1 sec (2) \u6839\u64da\u672c\u5c08\u6848\u7cfb\u7d71\u4e2d\u8a9e\u97f3\u8cc7\u6599\u53d6\u6a23\u6642\u6240\u63a1\u53d6\u7684\u97f3\u6846\u64f7\u53d6\u65b9\u5f0f\uff0c100 \u500b\u97f3\u6846\u7d04\u7b49\u65bc 1 sec \u9700\u6642\u9593\u4e5f\u6703\u589e\u52a0\u3002\u82e5\u662f\u80cc\u666f\u8a9e\u8005\u7684\u4eba\u6578\u9078\u53d6\u4f4e\u81f3 10 \u4eba\uff0c\u5247\u4ee5 Bhattacharyya \u8ddd\u96e2(\u8868\u4e09\u4e2d \u9032\u884c\u9a57\u8b49\u6e2c\u8a66\u3002 Frame \u4e2d\u6240\u8a2d\u5b9a\u7684\u97f3\u6846\u6578)\uff0c\u5247\u4ee5\u8a72\u7b46\u6e2c\u8a66\u8a9e\u97f3\u8cc7\u6599\u4e2d\u5168\u90e8\u7684\u8cc7\u6599\u4f86\u9032\u884c\u9a57\u8b49\u6e2c\u8a66\u3002 \u8005\u4eba\u6578\u6108\u591a\u6108\u80fd\u5920\u589e\u52a0\u53cd\u8a9e\u8005\u6a21\u578b\u7684\u9451\u5225\u5ea6\uff0c\u63d0\u5347\u9a57\u8b49\u4e4b\u6548\u80fd\uff0c\u4f46\u662f\u76f8\u5c0d\u5730\u8a08\u7b97\u91cf\u53ca\u6240 \u8868\u56db\u3001\u662f\u5426\u4f7f\u7528\u6027\u5225\u76f8\u95dc\u7684\u80cc\u666f\u8a9e\u8005\u6a21\u578b \u80cc\u666f\u8a9e\u8005\u4eba\u6578 \u6027\u5225\u76f8\u95dc \u5168\u90e8\u8a9e\u8005 10 0.9132 20 0.9504 30 0.9632 40 0.9753 50 0.9875 \u6846\u6578\u3002\u82e5\u8a72\u7b46\u6e2c\u8a66\u8a9e\u97f3\u8cc7\u6599\u7684\u97f3\u6846\u6578\u4e0d\u8db3(\u6e2c\u8a66\u8a9e\u97f3\u8cc7\u6599\u7684\u97f3\u6846\u7e3d\u6578 \uff1c Testing Data (\u6e2c\u8a66\u8a9e\u97f3\u8cc7\u6599\u7684\u97f3\u6846\u7e3d\u6578 \uff1c 100 \u500b\u97f3\u6846)\uff0c\u5247 \u4ee5\u8a72\u7b46\u6e2c\u8a66\u8a9e\u97f3\u8cc7\u6599\u4e2d\u5168\u90e8\u7684\u8cc7\u6599\u4f86 0.9716 (1) Testing Data Frame \u4e2d\u7684\u6578\u503c\u6240\u4ee3\u8868\u7684\u662f\u65bc\u6bcf\u4e00\u7b46\u6e2c\u8a66\u8a9e\u97f3\u8cc7\u6599\u4e2d\u5f9e\u982d\u958b\u59cb\u64f7\u53d6\u7684\u97f3 (2) \u6bcf\u4e00\u7b46\u6e2c\u8a66\u8a9e\u97f3\u8cc7\u6599\u5f9e\u982d\u958b\u59cb\u64f7\u53d6 100 \u500b\u97f3\u6846\u3002\u82e5\u8a72\u7b46\u6e2c\u8a66\u8a9e\u97f3\u8cc7\u6599\u7684\u97f3\u6846\u6578\u4e0d\u8db3 0.9582 \u8a3b\uff1a \u8a13\u7df4\u5176 GMM \u6a21\u578b\u3002 0.9556 300 0.9883 0.9854 0.9868 \uff1c Training Data Frame \u4e2d\u6240\u8a2d\u5b9a\u7684\u97f3\u6846\u6578)\uff0c\u5247 \u4ee5\u8a72\u4f4d\u8a9e\u8005\u6240\u8a3b\u518a\u5168\u90e8\u7684\u8a9e\u97f3\u8cc7\u6599\u4f86 0.9398 250 0.9868 0.9839 \u64f7\u53d6\u7684\u97f3\u6846\u6578\u3002\u82e5\u8a72\u4f4d\u8a9e\u8005\u8a3b\u518a\u7684\u8a9e\u97f3\u8cc7\u6599\u97f3\u6846\u6578\u4e0d\u8db3(\u8a3b\u518a\u8a9e\u97f3\u8cc7\u6599\u7684\u97f3\u6846\u7e3d\u6578 0.9853 0.9038 100 0.9765 0.9808 0.9787 150 0.9853 0.9824 0.9839 200 0.9897 0.9839 0.9868 18000 0.9765 0.9808 0.9787 \u672c\u7814\u7a76\u627f\u8499\u570b\u79d1\u6703\u5c08\u984c\u7814\u7a76\u8a08\u756b\u300c\u70ba\u96f2\u7aef\u670d\u52d9\u800c\u8a2d\u8a08\u4e4b\u667a\u6167\u7d42\u7aef\u61c9\u7528\u5b89\u5168\u5957\u4ef6\u300d\u7684\u90e8\u4efd \u8a3b\uff1a (1) Training Data Frame \u4e2d\u7684\u6578\u503c\u6240\u4ee3\u8868\u7684\u662f\u6bcf\u4e00\u4f4d\u8a9e\u8005\u65bc\u5176\u8a3b\u518a\u8a9e\u97f3\u8cc7\u6599\u4e2d\u5f9e\u982d\u958b\u59cb \u7d93\u8cbb\u88dc\u52a9\uff0c\u65b9\u5f97\u4ee5\u5b8c\u6210\u672c\u7814\u7a76\uff0c\u8b39\u6b64\u81f4\u8b1d\u3002
", "html": null, "type_str": "table", "text": "Proceedings of the Twenty-Fifth Conference on Computational Linguistics and Speech Processing (ROCLING 2013) \u6a21\u7d44\u5373\u6642\u540c\u6b65\u50b3\u9001\u81f3\u9a57\u8b49\u4f3a\u670d\u7aef\u3002\u7aef\u9ede\u5075\u6e2c\u4e3b\u8981\u4f7f\u7528\u80fd\u91cf\u53ca\u904e\u96f6\u9ede\u7387\u4f5c\u70ba\u7aef\u9ede\u5075\u6e2c\u7684\u7279 \u5fb5\u3002\u4f3a\u670d\u7aef\u5728\u63a5\u6536\u5230\u8a9e\u97f3\u4e32\u6d41\u5f8c\u6703\u9032\u884c\u5373\u6642\u540c\u6b65\u7684\u7279\u5fb5\u8a08\u7b97\u4ee5\u53ca\u8fa8\u8b58\u641c\u5c0b\u3002\u5728\u8fa8\u8b58\u4f3a\u670d \u5668\u7684\u5be6\u4f5c\u4e0a\uff0c\u7531\u65bc\u8fa8\u8b58\u8207\u9a57\u8b49\u8a08\u7b97\u8907\u96dc\u5ea6\u8f03\u9ad8\uff0c\u6240\u4ee5\u7279\u5fb5\u64f7\u53d6\u8207\u8fa8\u8b58\u7684\u6838\u5fc3\u662f\u4ee5 C++ \u5be6\u4f5c\uff0c\u4ee5\u9054\u5230\u9ad8\u6548\u7387\u7684\u5373\u6642\u8fa8\u8b58\u3002\u800c Socket \u63a5\u6536\u8a9e\u97f3\u4e32\u6d41\u7684\u90e8\u5206\u5247\u662f\u4ee5 Java \u8a9e\u8a00\u5be6\u4f5c\uff0c \u900f\u904e Java \u539f\u751f\u4ecb\u9762(Java Native Interface, JNI)\u898f\u7bc4\u547c\u53eb C++\u7de8\u8b6f\u4e4b\u9a57\u8b49\u5f15\u64ce\u539f\u751f\u78bc\u3002 Proceedings of the Twenty-Fifth Conference on Computational Linguistics and Speech Processing (ROCLING 2013) \u6a21\u578b\u4e4b\u6df7\u5408\u6578\u6108\u591a\u6642\uff0c\u5176 F \u503c\u4e5f\u6703\u96a8\u4e4b\u4e0a\u5347\u3002\u7531\u6b64\u6211\u5011\u53ef\u4ee5\u5f97\u77e5\uff1a\u6df7\u5408\u6578\u6108\u591a\u6108\u80fd\u63cf\u8ff0 \u51fa\u8a9e\u8005\u7684\u767c\u8072\u7279\u6027\uff0c\u4f46\u662f\u76f8\u5c0d\u5730\u8a08\u7b97\u91cf\u4ee5\u53ca\u6240\u9700\u6642\u9593\u4e5f\u6703\u589e\u52a0\uff1b\u6df7\u5408\u6578\u8f03\u5c11\u6642\uff0c\u7cfb\u7d71\u6548 \u80fd\u5c31\u6703\u4e0b\u964d\u3002\u4f8b\u5982\u6df7\u5408\u6578\u70ba 5 \u6642\uff0c\u5176 F \u503c\u50c5\u6709 0.9091\uff1b\u82e5\u6df7\u5408\u6578\u5230\u9054 15 \u6642\uff0c\u5247 F \u503c\u53ef \u4ee5\u5230\u9054 0.9868\uff1b\u800c\u7576\u6df7\u5408\u6578\u8d85\u904e 15 \u6642\uff0c\u7cfb\u7d71\u6548\u80fd\u5247\u5448\u73fe\u98fd\u548c\u72c0\u614b\u3002\u96d6\u7136\u7576\u6df7\u5408\u6578\u5230\u9054 60 \u6642\uff0c\u5176 F \u503c\u53ef\u4ee5\u5230\u9054 0.9993\uff0c\u4f46\u662f\u56e0\u70ba\u6211\u5011\u4f7f\u7528\u4e86 102 \u4f4d\u80cc\u666f\u8a9e\u8005\u4f86\u505a\u6e2c\u8a66\uff0c\u56e0\u6b64 \u82e5\u4f7f\u7528 60 \u500b Mixture \u6703\u4f7f\u5f97\u5c0d\u9ad8\u65af\u6df7\u5408\u6a21\u578b\u8a08\u7b97 Likelihood \u6642\u7684\u8a08\u7b97\u91cf\u9054\u5230 6120 \u500b\u9ad8 Proceedings of the Twenty-Fifth Conference on Computational Linguistics and Speech Processing (ROCLING 2013)" } } } }