{ "paper_id": "O13-1004", "header": { "generated_with": "S2ORC 1.0.0", "date_generated": "2023-01-19T08:03:56.082310Z" }, "title": "Quickly Personalizable Mobile Digit Speech Recognition System Based on Sphinx", "authors": [ { "first": "\u984f\u5b97\u8283", "middle": [], "last": "Tsung-Peng", "suffix": "", "affiliation": { "laboratory": "", "institution": "National Sun Yat-Sen University", "location": {} }, "email": "" }, { "first": "Yen", "middle": [], "last": "\u9673\u5609\u5e73", "suffix": "", "affiliation": { "laboratory": "", "institution": "National Sun Yat-Sen University", "location": {} }, "email": "" }, { "first": "Chia-Ping", "middle": [], "last": "Chen", "suffix": "", "affiliation": { "laboratory": "", "institution": "National Sun Yat-Sen University", "location": {} }, "email": "cpchen@cse.nsysu.edu.tw" } ], "year": "", "venue": null, "identifiers": {}, "abstract": "In this paper, we introduce a system for on-line digit speech recognition services. Besides the speech recognition service in our system, we also provide adaptation function to improve the noise-robustness between different environment. In the case of English digit recognition, our recognition system can achieve over 80% accuracy for a specific speaker by using a few adaptation data. We use Sphinx-4 as a speech recognition kernel in our system. Because Sphinx-4 is a system prepared exclusively for researchers, it is a flexible, modular and pluggable framework. We provide our experiment results on AURORA2, EAT and Android device recording. We use AURORA2 database training models that adapt by EAT and Android device recording. The experimental results show we can get high accuracy after a few adaptation.", "pdf_parse": { "paper_id": "O13-1004", "_pdf_hash": "", "abstract": [ { "text": "In this paper, we introduce a system for on-line digit speech recognition services. Besides the speech recognition service in our system, we also provide adaptation function to improve the noise-robustness between different environment. In the case of English digit recognition, our recognition system can achieve over 80% accuracy for a specific speaker by using a few adaptation data. We use Sphinx-4 as a speech recognition kernel in our system. Because Sphinx-4 is a system prepared exclusively for researchers, it is a flexible, modular and pluggable framework. We provide our experiment results on AURORA2, EAT and Android device recording. We use AURORA2 database training models that adapt by EAT and Android device recording. The experimental results show we can get high accuracy after a few adaptation.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Abstract", "sec_num": null } ], "body_text": [ { "text": "\u4e00 \u4e00 \u4e00\u3001 \u3001 \u3001\u7814 \u7814 \u7814\u7a76 \u7a76 \u7a76\u80cc \u80cc \u80cc\u666f \u666f \u666f\u3001 \u3001 \u3001\u52d5 \u52d5 \u52d5\u6a5f \u6a5f \u6a5f \u62dc\u79d1\u6280\u7684\u6f14\u9032\u53ca\u7db2\u8def\u767c\u5c55\u6240\u8cdc\uff0c\u8a9e\u97f3\u8fa8\u8b58\u6210\u70ba\u4e86\u751f\u6d3b\u4e0a\u65e5\u6f38\u91cd\u8981\u7684\u89d2\u8272\u5982 Google voice search [1] ", "cite_spans": [ { "start": 99, "end": 102, "text": "[1]", "ref_id": "BIBREF0" } ], "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": "\u8868 2\u3001 NOISE1 \u8a9e\u6599\u5eab\u9304\u88fd\u74b0\u5883\u6709\u95dc\u9589\u6240\u6709\u5bb6\u96fb\u8207\u9580\u7a97\u7684\u5b89\u975c\u5bbf\u820d \u4e2d(dormitory) \u3001 \u5b89 \u975c \u5bbf \u820d \u65bc \u958b \u555f \u7684 \u96fb \u98a8 \u6247 \u65c1 (fan) \u3001 \u4e0b \u73ed \u6642 \u6bb5 \u897f \u5b50 \u7063 \u6377 \u904b \u4e8c \u865f \u51fa \u53e3 \u65c1 \u7684 \u516c \u8eca \u7b49 \u5019 \u4ead (road) \u3001\u4e2d\u5c71\u5927\u5b78\u4e0b\u5348\u4e94\u9ede\u7684\u7c43\u7403\u5834 (basketball) \u3001\u4e2d\u5c71\u5927\u5b78\u4e2d\u5348 11 \u9ede\u534a L \u578b\u505c\u8eca \u5834 (parkingLot) \u53ca\u4e2d\u5c71\u5927\u5b78\u96fb\u8cc7\u5927\u6a13 F5017b \u5be6\u9a57\u5ba4 (laboratory)\u3002", "eq_num": "(" } ], "section": "", "sec_num": null }, { "text": "\u5206\u5225\u5c0d AURORA2 \u7684\u4e7e\u6de8\u8a13\u7df4\u8a9e\u6599\u8207\u591a\u74b0\u5883\u8a13\u7df4\u8a9e\u6599\u4f7f\u7528 SphinxTrain \u5f97\u5230\u4e7e\u6de8 \u8a9e\u6599 (clean) \u53ca\u591a\u74b0\u5883\u8a9e\u6599 (multi) \u5169\u500b\u8072\u5b78\u6a21\u578b\uff0c\u4f7f\u7528 AURORA2 \u7684\u6e2c\u8a66\u8a9e\u6599\u6240\u5f97\u5230 \u7684\u8fa8\u8b58\u7387\u5982\u8868 5 \u6240\u793a\uff0c\u672c\u8ad6\u6587\u4f7f\u7528\u5b57\u6a21\u578b (word dependant model) \u4f5c\u70ba\u57fa\u790e\u6a21\u578b\uff0c\u628a \u76f8\u540c\u767c\u97f3\u7684\u97f3\u7d20\u6a21\u578b\u5171\u7528\u6240\u5f97\u5230\u7684\u7d50\u679c\u4e5f\u975e\u5e38\u76f8\u8fd1\u65bc\u73fe\u5728\u7684\u7d50\u679c\u3002\u5be6\u9a57\u4e2d\u5229\u7528 EAT DIGIT \u8207 NOISE1 \u8a9e\u6599\u505a\u4e00\u7cfb\u5217\u7684\u8abf\u9069\u5be6\u9a57\uff0c\u5728\u9019\u4e9b\u8abf\u9069\u5be6\u9a57\u4e2d\u6240\u7528\u5230\u7684\u8abf\u9069\u6cd5\u70ba\u6700 \u5927\u5f8c\u9a57 (Maximum A Posteriori, MAP) [17] \u65b9\u6cd5\u4f86\u9032\u884c\u3002 (\u4e09 \u4e09 \u4e09)\u3001 \u3001 \u3001AURORA2 \u8072 \u8072 \u8072\u5b78 \u5b78 \u5b78\u6a21 \u6a21 \u6a21\u578b \u578b \u578b\u4f7f \u4f7f \u4f7f\u7528 \u7528 \u7528 EAT DIGIT \u8a9e \u8a9e \u8a9e\u6599 \u6599 \u6599\u5eab \u5eab \u5eab\u8a9e \u8a9e \u8a9e\u6599 \u6599 \u6599\u8abf \u8abf \u8abf\u9069 \u9069 \u9069\u5be6 \u5be6 \u5be6\u9a57 \u9a57 \u9a57 \u70ba\u4e86\u89e3\u5916\u570b\u8a9e\u8a00\u8154\u8abf\u5728\u53d7\u904e\u8a13\u7df4\u5f8c\u8207\u672a\u53d7\u904e\u8a13\u7df4\u7684\u5dee\u7570\u9032\u884c\u82f1\u8a9e\u7cfb\u8207\u975e\u82f1\u8a9e\u7cfb\u5b78 \u751f\u7684\u8154\u8abf\u6bd4\u8f03\u3001\u8abf\u9069\u6548\u679c\u8207\u53e5\u6578\u95dc\u4fc2\u3001\u8de8\u74b0\u5883\u8abf\u9069\u6548\u679c\u4e09\u9805\u5be6\u9a57\u3002EAT DIGIT \u4e00\u5171 \u6709 gsm\u3001pstn\u3001mic16k \u4e09\u7a2e\u9304\u97f3\u65b9\u5f0f\uff0c\u518d\u7d30\u5206\u70ba gsm \u82f1\u8a9e\u7cfb\u5b78\u751f\u7684\u8a9e\u6599 (gsmE) \u3001gsm \u975e \u82f1\u8a9e\u7cfb\u5b78\u751f\u7684\u8a9e\u6599 (gsmN)\u3001pstnE\u3001pstnN\u3001mic16kE\u3001mic16kN\uff0c\u6700\u5f8c\u628a\u9019\u516d\u7a2e\u689d\u4ef6\u7684 \u8a9e\u6599\u5206\u6210\u6e2c\u8a66\u8a9e\u6599\u4ee5\u53ca\u8abf\u9069\u8a9e\u6599\uff0c\u5982\u8868 6 \u6240\u793a\u3002\u628a\u6bcf\u4e00\u7a2e\u689d\u4ef6\u7684\u8a9e\u6599\u518d\u5404\u5206\u6210\u5169\u534a\uff0c \u4e00\u534a\u505a\u6e2c\u8a66\u8a9e\u6599\u4e00\u534a\u505a\u8abf\u9069\u8a9e\u6599\u3002\u5206\u5225\u5728\u5f8c\u9762\u52a0\u4e0a t \u8207 a \u4ee3\u8868\u6e2c\u8a66\u8a9e\u6599\u53ca\u8abf\u9069\u8a9e \u6599\uff0c\u5c07\u6bcf\u4e00\u7a2e\u74b0\u5883\u7684\u8a9e\u6599\u4e00\u5171\u5206\u6210\u56db\u4efd (E test\u3001E adapt\u3001NE test\u3001NE adapt) \u3002 1\u3001 \u3001 \u3001EAT DIGIT \u82f1 \u82f1 \u82f1\u8a9e \u8a9e \u8a9e\u7cfb \u7cfb \u7cfb\u8207 \u8207 \u8207\u975e \u975e \u975e\u82f1 \u82f1 \u82f1\u8a9e \u8a9e \u8a9e\u7cfb \u7cfb \u7cfb\u8154 \u8154 \u8154\u8abf \u8abf \u8abf\u6bd4 \u6bd4 \u6bd4\u8f03 \u8f03 \u8f03 \u5229\u7528\u5404\u74b0\u5883 E a \u53ca N a \u7684\u90e8\u5206\u8abf\u9069\u6210\u82f1\u8a9e\u7cfb\u6a21\u578b\u53ca\u975e\u82f1\u8a9e\u7cfb\u6a21\u578b\uff0c\u518d\u4f7f\u7528 E t \u8207 N t \u7684", "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": "Your word is my command\": Google search by voice: A case study", "authors": [ { "first": "J", "middle": [], "last": "Schalkwyk", "suffix": "" }, { "first": "D", "middle": [], "last": "Beeferman", "suffix": "" }, { "first": "F", "middle": [], "last": "Beaufays", "suffix": "" }, { "first": "B", "middle": [], "last": "Byrne", "suffix": "" }, { "first": "C", "middle": [], "last": "Chelba", "suffix": "" }, { "first": "M", "middle": [], "last": "Cohen", "suffix": "" }, { "first": "M", "middle": [], "last": "Kamvar", "suffix": "" }, { "first": "B", "middle": [], "last": "Strope", "suffix": "" } ], "year": 2010, "venue": "Advances in Speech Recognition: Mobile Environments", "volume": "", "issue": "", "pages": "61--90", "other_ids": {}, "num": null, "urls": [], "raw_text": "J. Schalkwyk, D. Beeferman, F. Beaufays, B. Byrne, C. Chelba, M. Cohen, M. Kamvar, and B. Strope, \"\"Your word is my command\": Google search by voice: A case study,\" in Advances in Speech Recognition: Mobile Environments, Call Centers and Clinics, 2010, ch. 4, pp. 61-90.", "links": null }, "BIBREF1": { "ref_id": "b1", "title": "Comparison of voice search applications on ios", "authors": [ { "first": "I", "middle": [], "last": "Vanduyn", "suffix": "" } ], "year": null, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "I. VanDuyn, \"Comparison of voice search applications on ios,\" http://www.isaacvanduyn. com/downloads/research-proposal.pdf, [Online]. Available.", "links": null }, "BIBREF2": { "ref_id": "b2", "title": "A large scale study of wireless search behavior: Google mobile search", "authors": [ { "first": "M", "middle": [], "last": "Kamvar", "suffix": "" }, { "first": "S", "middle": [], "last": "Baluja", "suffix": "" } ], "year": 2006, "venue": "Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, ser. CHI '06", "volume": "", "issue": "", "pages": "701--709", "other_ids": {}, "num": null, "urls": [], "raw_text": "M. Kamvar and S. Baluja, \"A large scale study of wireless search behavior: Google mo- bile search,\" in Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, ser. CHI '06. New York, NY, USA: ACM, 2006, pp. 701-709.", "links": null }, "BIBREF3": { "ref_id": "b3", "title": "Cyberon voice commander \u591a\u570b\u8a9e\u8a00\u8a9e\u97f3\u547d\u4ee4\u7cfb\u7d71 (Cyberon Voice Commander -a Multilingual Voice Command System)", "authors": [ { "first": "T", "middle": [ "X" ], "last": "He", "suffix": "" }, { "first": "J.-J", "middle": [], "last": "Liou", "suffix": "" } ], "year": 2007, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "T. X. He and J.-J. Liou, \"Cyberon voice commander \u591a\u570b\u8a9e\u8a00\u8a9e\u97f3\u547d\u4ee4\u7cfb\u7d71 (Cyberon Voice Commander -a Multilingual Voice Command System) [In Chinese],\" in ROCLING, 2007.", "links": null }, "BIBREF4": { "ref_id": "b4", "title": "Voice based control for humanoid teleoperation", "authors": [ { "first": "Y", "middle": [], "last": "Lu", "suffix": "" }, { "first": "L", "middle": [], "last": "Liu", "suffix": "" }, { "first": "S", "middle": [], "last": "Chen", "suffix": "" }, { "first": "Q", "middle": [], "last": "Huang", "suffix": "" } ], "year": 2010, "venue": "Intelligent System Design and Engineering Application (ISDEA), 2010 International Conference on", "volume": "2", "issue": "", "pages": "814--818", "other_ids": {}, "num": null, "urls": [], "raw_text": "Y. Lu, L. Liu, S. Chen, and Q. Huang, \"Voice based control for humanoid teleoperation,\" in Intelligent System Design and Engineering Application (ISDEA), 2010 International Conference on, vol. 2, 2010, pp. 814-818.", "links": null }, "BIBREF5": { "ref_id": "b5", "title": "Above the clouds: A berkeley view of cloud computing", "authors": [ { "first": "M", "middle": [], "last": "Armbrust", "suffix": "" }, { "first": "A", "middle": [], "last": "Fox", "suffix": "" }, { "first": "R", "middle": [], "last": "Griffith", "suffix": "" }, { "first": "A", "middle": [ "D" ], "last": "Joseph", "suffix": "" }, { "first": "R", "middle": [ "H" ], "last": "Katz", "suffix": "" }, { "first": "A", "middle": [], "last": "Konwinski", "suffix": "" }, { "first": "G", "middle": [], "last": "Lee", "suffix": "" }, { "first": "D", "middle": [ "A" ], "last": "Patterson", "suffix": "" }, { "first": "A", "middle": [], "last": "Rabkin", "suffix": "" }, { "first": "I", "middle": [], "last": "Stoica", "suffix": "" }, { "first": "M", "middle": [], "last": "Zaharia", "suffix": "" } ], "year": 2009, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "M. Armbrust, A. Fox, R. Griffith, A. D. Joseph, R. H. Katz, A. Konwinski, G. Lee, D. A. Patterson, A. Rabkin, I. Stoica, and M. Zaharia, \"Above the clouds: A berkeley view of cloud computing,\" EECS Department, University of California, Berkeley, Tech. Rep. UCB/EECS-2009-28, Feb 2009.", "links": null }, "BIBREF6": { "ref_id": "b6", "title": "Speech browsing the world wide web", "authors": [ { "first": "J", "middle": [], "last": "Borges", "suffix": "" }, { "first": "J", "middle": [], "last": "Jimenez", "suffix": "" }, { "first": "N", "middle": [], "last": "Rodriquez", "suffix": "" } ], "year": 1999, "venue": "IEEE SMC '99 Conference Proceedings. 1999 IEEE International Conference on", "volume": "4", "issue": "", "pages": "80--86", "other_ids": {}, "num": null, "urls": [], "raw_text": "J. Borges, J. Jimenez, and N. Rodriquez, \"Speech browsing the world wide web,\" in Systems, Man, and Cybernetics, 1999. IEEE SMC '99 Conference Proceedings. 1999 IEEE International Conference on, vol. 4, 1999, pp. 80-86 vol.4.", "links": null }, "BIBREF7": { "ref_id": "b7", "title": "An introduction to hidden markov models", "authors": [ { "first": "L", "middle": [], "last": "Rabiner", "suffix": "" }, { "first": "B.-H", "middle": [], "last": "Juang", "suffix": "" } ], "year": 1986, "venue": "ASSP Magazine, IEEE", "volume": "3", "issue": "1", "pages": "4--16", "other_ids": {}, "num": null, "urls": [], "raw_text": "L. Rabiner and B.-H. Juang, \"An introduction to hidden markov models,\" ASSP Magazine, IEEE, vol. 3, no. 1, pp. 4-16, 1986.", "links": null }, "BIBREF8": { "ref_id": "b8", "title": "Stranded gaussian mixture hidden markov models for robust speech recognition", "authors": [ { "first": "Y", "middle": [], "last": "Zhao", "suffix": "" }, { "first": "B.-H", "middle": [], "last": "Juang", "suffix": "" } ], "year": 2012, "venue": "Acoustics, Speech and Signal Processing (ICASSP), 2012 IEEE International Conference on", "volume": "", "issue": "", "pages": "4301--4304", "other_ids": {}, "num": null, "urls": [], "raw_text": "Y. Zhao and B.-H. Juang, \"Stranded gaussian mixture hidden markov models for robust speech recognition,\" in Acoustics, Speech and Signal Processing (ICASSP), 2012 IEEE International Conference on, 2012, pp. 4301-4304.", "links": null }, "BIBREF9": { "ref_id": "b9", "title": "Exploiting sparsity in stranded hidden markov models for automatic speech recognition", "authors": [], "year": 2012, "venue": "Signals, Systems and Computers (ASILOMAR), 2012 Conference Record of the Forty Sixth Asilomar Conference on", "volume": "", "issue": "", "pages": "1623--1625", "other_ids": {}, "num": null, "urls": [], "raw_text": "--, \"Exploiting sparsity in stranded hidden markov models for automatic speech recog- nition,\" in Signals, Systems and Computers (ASILOMAR), 2012 Conference Record of the Forty Sixth Asilomar Conference on, 2012, pp. 1623-1625.", "links": null }, "BIBREF10": { "ref_id": "b10", "title": "Multilingual acoustic modeling for speech recognition based on subspace gaussian mixture models", "authors": [ { "first": "L", "middle": [], "last": "Burget", "suffix": "" }, { "first": "P", "middle": [], "last": "Schwarz", "suffix": "" }, { "first": "M", "middle": [], "last": "Agarwal", "suffix": "" }, { "first": "P", "middle": [], "last": "Akyazi", "suffix": "" }, { "first": "K", "middle": [], "last": "Feng", "suffix": "" }, { "first": "A", "middle": [], "last": "Ghoshal", "suffix": "" }, { "first": "O", "middle": [], "last": "Glembek", "suffix": "" }, { "first": "N", "middle": [], "last": "Goel", "suffix": "" }, { "first": "M", "middle": [], "last": "Karafiat", "suffix": "" }, { "first": "D", "middle": [], "last": "Povey", "suffix": "" }, { "first": "A", "middle": [], "last": "Rastrow", "suffix": "" }, { "first": "R", "middle": [], "last": "Rose", "suffix": "" }, { "first": "S", "middle": [], "last": "Thomas", "suffix": "" } ], "year": 2010, "venue": "Acoustics Speech and Signal Processing (ICASSP)", "volume": "", "issue": "", "pages": "4334--4337", "other_ids": {}, "num": null, "urls": [], "raw_text": "L. Burget, P. Schwarz, M. Agarwal, P. Akyazi, K. Feng, A. Ghoshal, O. Glembek, N. Goel, M. Karafiat, D. Povey, A. Rastrow, R. Rose, and S. Thomas, \"Multilingual acoustic modeling for speech recognition based on subspace gaussian mixture models,\" in Acoustics Speech and Signal Processing (ICASSP), 2010 IEEE International Confer- ence on, 2010, pp. 4334-4337.", "links": null }, "BIBREF11": { "ref_id": "b11", "title": "Subspace gaussian mixture models for speech recognition", "authors": [ { "first": "D", "middle": [], "last": "Povey", "suffix": "" }, { "first": "L", "middle": [], "last": "Burget", "suffix": "" }, { "first": "M", "middle": [], "last": "Agarwal", "suffix": "" }, { "first": "P", "middle": [], "last": "Akyazi", "suffix": "" }, { "first": "K", "middle": [], "last": "Feng", "suffix": "" }, { "first": "A", "middle": [], "last": "Ghoshal", "suffix": "" }, { "first": "O", "middle": [], "last": "Glembek", "suffix": "" }, { "first": "N", "middle": [], "last": "Goel", "suffix": "" }, { "first": "M", "middle": [], "last": "Karafiat", "suffix": "" }, { "first": "A", "middle": [], "last": "Rastrow", "suffix": "" }, { "first": "R", "middle": [], "last": "Rose", "suffix": "" }, { "first": "P", "middle": [], "last": "Schwarz", "suffix": "" }, { "first": "S", "middle": [], "last": "Thomas", "suffix": "" } ], "year": 2010, "venue": "Acoustics Speech and Signal Processing (ICASSP), 2010 IEEE International Conference on", "volume": "", "issue": "", "pages": "4330--4333", "other_ids": {}, "num": null, "urls": [], "raw_text": "D. Povey, L. Burget, M. Agarwal, P. Akyazi, K. Feng, A. Ghoshal, O. Glembek, N. Goel, M. Karafiat, A. Rastrow, R. Rose, P. Schwarz, and S. Thomas, \"Subspace gaussian mixture models for speech recognition,\" in Acoustics Speech and Signal Processing (ICASSP), 2010 IEEE International Conference on, 2010, pp. 4330-4333.", "links": null }, "BIBREF12": { "ref_id": "b12", "title": "The HTK Book Version 3.4", "authors": [ { "first": "S", "middle": [ "J" ], "last": "Young", "suffix": "" }, { "first": "D", "middle": [], "last": "Kershaw", "suffix": "" }, { "first": "J", "middle": [], "last": "Odell", "suffix": "" }, { "first": "D", "middle": [], "last": "Ollason", "suffix": "" }, { "first": "V", "middle": [], "last": "Valtchev", "suffix": "" }, { "first": "P", "middle": [], "last": "Woodland", "suffix": "" } ], "year": 2006, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "S. J. Young, D. Kershaw, J. Odell, D. Ollason, V. Valtchev, and P. Woodland, The HTK Book Version 3.4. Cambridge University Press, 2006.", "links": null }, "BIBREF13": { "ref_id": "b13", "title": "Sphinx-4: a flexible open source framework for speech recognition", "authors": [ { "first": "W", "middle": [], "last": "Walker", "suffix": "" }, { "first": "P", "middle": [], "last": "Lamere", "suffix": "" }, { "first": "P", "middle": [], "last": "Kwok", "suffix": "" }, { "first": "B", "middle": [], "last": "Raj", "suffix": "" }, { "first": "R", "middle": [], "last": "Singh", "suffix": "" }, { "first": "E", "middle": [], "last": "Gouvea", "suffix": "" }, { "first": "P", "middle": [], "last": "Wolf", "suffix": "" }, { "first": "J", "middle": [], "last": "Woelfel", "suffix": "" } ], "year": 2004, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "W. Walker, P. Lamere, P. Kwok, B. Raj, R. Singh, E. Gouvea, P. Wolf, and J. Woelfel, \"Sphinx-4: a flexible open source framework for speech recognition,\" Mountain View, CA, USA, Tech. Rep., 2004.", "links": null }, "BIBREF14": { "ref_id": "b14", "title": "The aurora experimental framework for the performance evaluation of speech recognition systems under noisy conditions", "authors": [ { "first": "D", "middle": [], "last": "Pearce", "suffix": "" }, { "first": "H", "middle": [], "last": "Hirsch", "suffix": "" }, { "first": "E", "middle": [ "E D" ], "last": "Gmbh", "suffix": "" } ], "year": 2000, "venue": "ISCA ITRW ASR2000", "volume": "", "issue": "", "pages": "29--32", "other_ids": {}, "num": null, "urls": [], "raw_text": "D. Pearce, H. g\u00fcnter Hirsch, and E. E. D. Gmbh, \"The aurora experimental framework for the performance evaluation of speech recognition systems under noisy conditions,\" in in ISCA ITRW ASR2000, 2000, pp. 29-32.", "links": null }, "BIBREF15": { "ref_id": "b15", "title": "\u53f0\u7063\u53e3\u97f3\u82f1\u8a9e\u8a9e\u6599\u5eab\u8aaa\u660e English Across Taiwan (EAT)", "authors": [ { "first": "", "middle": [], "last": "\u4e2d\u83ef\u6c11\u570b\u8a08\u7b97\u8a9e\u8a00\u5b78\u5b78\u6703", "suffix": "" } ], "year": null, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "\u4e2d\u83ef\u6c11\u570b\u8a08\u7b97\u8a9e\u8a00\u5b78\u5b78\u6703, \"\u53f0\u7063\u53e3\u97f3\u82f1\u8a9e\u8a9e\u6599\u5eab\u8aaa\u660e English Across Taiwan (EAT),\" http://www.aclclp.org.tw/doc/eat brief.pdf, [Online]. Available.", "links": null }, "BIBREF16": { "ref_id": "b16", "title": "Speaker adaptation based on map estimation of hmm parameters", "authors": [ { "first": "C.-H", "middle": [], "last": "Lee", "suffix": "" }, { "first": "J.-L", "middle": [], "last": "Gauvain", "suffix": "" } ], "year": 1993, "venue": "Proceedings of the 1993 IEEE international conference on Acoustics, speech, and signal processing: speech processing", "volume": "II", "issue": "", "pages": "558--561", "other_ids": {}, "num": null, "urls": [], "raw_text": "C.-H. Lee and J.-L. Gauvain, \"Speaker adaptation based on map estimation of hmm parameters,\" in Proceedings of the 1993 IEEE international conference on Acoustics, speech, and signal processing: speech processing -Volume II, ser. ICASSP'93. Wash- ington, DC, USA: IEEE Computer Society, 1993, pp. 558-561.", "links": null } }, "ref_entries": { "TABREF0": { "type_str": "table", "content": "
\u56de\u61c9 \u7db2 \u969b \u7db2 \u8def \u9700\u6c42 \u73fe\u5728\u5e38\u898b\u5230\u7684\u4e3b\u6d41\u7684\u8a9e\u97f3\u8fa8\u8b58\u7814\u7a76\u5927\u591a\u662f\u4ee5\u96b1\u85cf\u5f0f\u99ac\u53ef\u592b\u6a21\u578b (Hidden Markov \u8fa8\u8b58\u7d50\u679c \u8a9e\u6599\u5eab Sphinx-4 SphinxTrain \u4f7f \u7528 \u8005 \u4ecb \u9762 \u4f7f\u7528\u8005 \u8a9e\u6599 \u8072\u5b78\u6a21\u578b \u4f3a \u670d \u7aef \u61c9 \u7528 \u7a0b \u5f0f \u7528 \u6236 \u7aef \u61c9 \u7528 \u7a0b \u5f0f \u8a13\u7df4 \u8abf\u9069 \u7522\u751f \u8fa8\u8b58 \u6536\u96c6\u8a9e\u6599 \u547d\u4ee4 \u8a9e\u6599 \u63a7\u5236 \u63a7\u5236\u547d\u4ee4 \u5716 1\u3001 \u7cfb\u7d71\u67b6\u69cb\u5716 \u7528\u6236\u7aef\u53ef\u80fd\u662f\u4e00\u53f0\u500b\u4eba\u96fb\u8166\u6216\u5c0f\u578b\u5de5\u4f5c\u7ad9\uff0c\u672c\u8eab\u5c31\u5177\u5099\u5b8c\u6574\u7368\u7acb\u4f5c\u696d\u7684\u80fd\u529b\uff1b\u4f3a\u670d\u7aef \u5247\u662f\u4e00\u53f0\u8f03\u5927\u578b\u7684\u4f3a\u670d\u5668\u6216\u96fb\u8166\u4e3b\u6a5f\uff0c\u800c\u5728\u7528\u6236\u7aef\u53ca\u4f3a\u670d\u7aef\u4e4b\u9593\u5247\u85c9\u8457\u53ef\u9760\u7684\u901a\u4fe1\u5354 \u5b9a\u9023\u7d50\u3002 \u672c\u7cfb\u7d71\u4ee5 HTTP (HyperText Transfer Protocol)\u7684\u65b9\u5f0f\u5efa\u7acb\u4e3b\u5f9e\u5f0f\u67b6\u69cb\uff0c\u4e00\u500b\u4f3a\u670d \u7aef (server) \u900f\u904e\u7db2\u8def\u4f86\u540c\u6642\u670d\u52d9\u591a\u500b\u7528\u6236\u7aef (client)\uff0cHTTP \u662f\u7db2\u969b\u7db2\u8def\u61c9\u7528\u6700\u70ba\u5ee3\u6cdb \u7684\u4e00\u7a2e\u7db2\u8def\u5354\u5b9a\uff0c\u5b83\u7684\u597d\u8655\u5728\u65bc\u80fd\u5920\u5bb9\u6613\u7684\u4f7f\u7528\u7db2\u9801\u4f3a\u670d\u5668\u67b6\u69cb\u51fa\u7528\u6236\u7aef\u7d66\u700f\u89bd\u5668\u4f7f \u7528\uff0c\u800c\u4e14\u5728\u5176\u5b83\u88dd\u7f6e\u4e0a\u4e5f\u5f88\u5bb9\u6613\u80fd\u5920\u8a2d\u8a08\u51fa\u7b26\u5408\u689d\u4ef6\u7684\u7528\u6236\u7aef\uff0c\u5716 1 \u8868\u793a\u4e86\u6574\u500b\u7cfb\u7d71 \u67b6\u69cb\uff0c\u7528\u6236\u7aef\u8207\u4f3a\u670d\u7aef\u5206\u5225\u4f7f\u7528\u4e0d\u540c\u7684\u61c9\u7528\u7a0b\u5f0f\u4f86\u63a7\u5236\uff0c\u4f7f\u7528\u8005\u900f\u904e\u4f7f\u7528\u8005\u4ecb\u9762 (user interface) \u8207\u7528\u6236\u7aef\u61c9\u7528\u7a0b\u5f0f\u6e9d\u901a\uff0c\u7528\u6236\u7aef\u61c9\u7528\u7a0b\u5f0f\u5c07\u8a9e\u6599\u53ca\u547d\u4ee4\u4ee5\u9700\u6c42\u7684\u65b9\u5f0f\u9001\u51fa\u81f3 \u4f3a\u670d\u7aef\uff0c\u4f3a\u670d\u7aef\u61c9\u7528\u7a0b\u5f0f\u6536\u5230\u9700\u6c42\u5f8c\u91dd\u5c0d\u6240\u9700\u63a7\u5236\u8fa8\u8b58\u5de5\u5177\u505a\u8fa8\u8b58\u6216\u8abf\u9069\u7684\u52d5\u4f5c\uff0c\u5b8c \u6210\u5f8c\u628a\u5c07\u8fa8\u8b58\u7d50\u679c\u6216\u5b8c\u6210\u8a0a\u606f\u56de\u61c9\u7d66\u7528\u6236\u7aef\u61c9\u7528\u7a0b\u5f0f\u4ee5\u64cd\u63a7\u4f7f\u7528\u8005\u4ecb\u9762\u3002 \u4e09 \u4e09 \u4e09\u3001 \u3001 \u3001\u5be6 \u5be6 \u5be6\u9a57 \u9a57 \u9a57 \u8a13\u7df4 AURORA2 \u6e2c\u8a66\u7d50\u679c \u8abf\u9069 \u6e2c\u8a66 \u8a13\u7df4\u8a9e\u6599 EAT DIGIT \u8abf\u9069\u8a9e\u6599 NOISE1 \u8abf\u9069\u8a9e\u6599 EAT Digit \u8abf\u9069\u6a21\u578b NOISE1 \u8abf\u9069\u6a21\u578b AURORA2 \u6e2c\u8a66\u8a9e\u6599 NOISE1 \u6e2c\u8a66\u8a9e\u6599 EAT DIGIT \u6e2c\u8a66\u7d50\u679c NOISE1 \u6e2c\u8a66\u7d50\u679c Proceedings AURORA2 \uff1a\u4f7f\u7528\u4e0d\u540c\u52a0\u6210\u6027\u566a\u97f3\u3001\u8a0a\u566a\u6bd4\u4f86\u6e2c\u8a66\u8a9e\u97f3\u8fa8\u8b58\u7cfb\u7d71\u5f37\u5065\u6027\u7684\u8a9e\u6599\u5eab\u3002 EAT DIGIT \uff1a\u5f9e EAT \u4e2d\u7684\u53ef\u7528\u8a9e\u6599\u4e2d\u904e\u6ffe\u51fa\u7d14\u82f1\u6587\u6578\u7528\u7684\u90e8\u5206\uff0c\u53e5\u6578\u5982\u8868 1\u3002 \u8868 1\u3001 \u53f0\u7063\u53e3\u97f3\u82f1\u8a9e\u8a9e\u6599\u5eab\u53ef\u7528\u7d14\u82f1\u6587\u6578\u5b57\u8a9e\u6599\u53e5\u6578 \u74b0\u5883\uff3c\u5206\u985e \u975e\u82f1\u8a9e\u7cfb\u5b78\u751f \u82f1\u8a9e\u7cfb\u5b78\u751f \u5973 \u7537 \u52a0\u7e3d \u5973 \u7537 \u52a0\u7e3d \u7e3d\u548c gsm 212 334 546 386 156 542 1088 pstn 168 171 439 341 136 477 916 mic16k 421 697 1118 794 318 1112 2230 NOISE1 \uff1a\u5f9e AURORA2 \u8a9e\u6599\u5eab\u4e2d\u9078\u64c7 NOISE1 \u7684\u6587\u672c (corpus) \u9304\u88fd(\u6545\u7a31 NOISE1 \u8a9e \u6599\u5eab)\uff0c\u4f7f\u7528\u88dd\u7f6e DHD (HTC Desire HD)\u53ca WFS(Wildfire S)\u9304\u88fd 8KHz 16bits PCM \u683c Model, \u4e3b\u5f9e\u5f0f\u67b6\u69cb\u662f\u4e00\u7a2e\u904b\u7528\u7db2\u8def\u6280\u8853\u3001\u958b\u653e\u7684\u67b6\u69cb\u4f86\u964d\u4f4e\u6210\u672c\u7684\u4e00\u7a2e\u5c0f\u578b\u5316\u96fb\u8166\u7cfb\u7d71\uff0c AURORA2 \u57fa\u790e\u6a21\u578b EAT DIGIT \u6e2c\u8a66\u8a9e\u6599 \u5f0f\u8a9e\u97f3\u6a94\u6848\uff0c\u6574\u500b\u8a9e\u6599\u5eab\u5982\u8868 2 \u6240\u793a\u3002
", "text": "\u3001Iphone Siri [2] \u53ca\u5176\u5b83\u76f8\u95dc\u61c9\u7528 [3] [4] [5]\uff0c\u884d\u751f\u4e86\u8a31\u591a\u53ef\u4ee5\u9023\u4e0a\u7db2\u969b\u7db2 \u8def\u7684\u79d1\u6280\u7522\u54c1 (\u5982 PDA \u3001\u667a\u6167\u578b\u624b\u6a5f\u3001\u5e73\u7248\u96fb\u8166)\u3002\u9019\u4e9b\u79d1\u6280\u7522\u54c1\u90fd\u5df2\u7d93\u6210\u70ba\u4e86\u73fe\u4ee3\u4eba \u7684\u751f\u6d3b\u5fc5\u9700\u54c1\uff0c\u4f46\u662f\u5927\u591a\u6578\u90fd\u662f\u4f7f\u7528\u50b3\u7d71\u7684\u6309\u9375\u4f86\u9032\u884c\u64cd\u4f5c\uff0c\u60f3\u8981\u5229\u7528\u6309\u9375\u9748\u6d3b\u7684\u64cd \u4f5c\u9019\u4e9b\u4e0d\u540c\u7684\u88dd\u7f6e\u662f\u975e\u5e38\u56f0\u96e3\u7684\u3002\u4f46\u5982\u679c\u6211\u5011\u7684\u88dd\u7f6e\u4e0d\u4fb7\u9650\u65bc\u6309\u9375\u8f38\u5165\u800c\u4f7f\u7528\u8a9e\u97f3\u8f38 \u5165\u4f86\u63a7\u5236\u9019\u4e9b\u88dd\u7f6e\uff0c\u751a\u81f3\u4e0d\u9700\u8981\u628a\u624b\u6a5f\u5f9e\u5305\u5305\u4e2d\u62ff\u51fa\u4f86\u5c31\u80fd\u5920\u64a5\u51fa\u96fb\u8a71\u8207\u670b\u53cb\u4ea4\u8ac7\u3002 \u628a\u8a9e\u97f3\u8b8a\u6210\u96a8\u8eab\u651c\u5e36\u7684\u842c\u7528\u9059\u63a7\u5668\u80fd\u5920\u5927\u5927\u7684\u6539\u5584\u4f7f\u7528\u4e0a\u7684\u4fbf\u5229\u6027\uff0c\u5373\u4f7f\u662f\u8eab\u9ad4\u6709\u6b98 \u7f3a\u7684\u4eba\u53ea\u9700\u8981\u900f\u904e\u53e3\u8a9e\uff0c\u4e5f\u80fd\u5229\u7528\u9019\u500b\u7cfb\u7d71\u4f86\u64cd\u4f5c\u9019\u4e9b\u73fe\u4ee3\u79d1\u6280\u7684\u624b\u6301\u88dd\u7f6e\u3002 \u73fe\u5728\u5927\u591a\u6578\u7684\u5373\u6642\u8a9e\u97f3\u8fa8\u8b58\u7cfb\u7d71\u90fd\u662f\u5efa\u7acb\u5728\u7db2\u969b\u7db2\u8def\u4e0a\uff0c\u5728\u8fa8\u8b58\u7684\u904e\u7a0b\u4e2d\u4f7f\u7528\u8005\u900f \u904e\u500b\u4eba\u96fb\u8166\u6216\u662f\u5176\u5b83\u88dd\u7f6e\u5c07\u8a9e\u97f3\u50b3\u81f3\u4f3a\u670d\u5668\u4e0a\uff0c\u5f85\u4f3a\u670d\u5668\u8fa8\u8b58\u5b8c\u6210\u5f8c\u5c07\u7d50\u679c\u56de\u50b3\uff0c\u628a \u8a9e\u97f3\u8fa8\u8b58\u76f8\u95dc\u7b49\u8f03\u8f03\u8017\u8cbb\u8cc7\u6e90\u7684\u5de5\u4f5c\u90fd\u4ea4\u7d66\u4f3a\u670d\u5668\u904b\u7b97\u3002\u9019\u7a2e\u67b6\u69cb\u8b93\u4f7f\u7528\u8005\u4e0d\u9700\u8981\u4f7f \u7528\u9ad8\u6548\u80fd\u7684\u88dd\u7f6e\u5c31\u80fd\u4f7f\u7528\u8a9e\u8fa8\u8b58\u7684\u670d\u52d9\uff0c\u50cf\u96f2\u7aef\u904b\u7b97\u670d\u52d9 [6] \u591a\u6578\u90fd\u5efa\u7acb\u65bc\u5927\u578b\u7684\u5206\u6563 \u5f0f\u4f3a\u670d\u5668\u4e0a\u3002\u5728 [7] \u4e2d\u63d0\u5230\uff0c\u4eba\u985e\u53ef\u7528\u8a9e\u97f3\u8f38\u5165\u4f86\u63a7\u5236\u700f\u89bd\u5668\u6307\u6a19\u4ee5\u589e\u9032\u4f7f\u7528\u8005\u8207\u7db2\u9801 \u7684\u4e92\u52d5\u3002 \u5728\u672c\u8ad6\u6587\u5c08\u6ce8\u5728\u5efa\u7acb\u4e00\u500b\u80fd\u5920\u517c\u5177\u670d\u52d9\u8207\u7814\u7a76\u7684\u8a9e\u97f3\u8fa8\u8b58\u7db2\u8def\u7cfb\u7d71\uff0c\u7814\u7a76\u4e0d\u540c\u53e3 \u97f3\u3001\u566a\u97f3\u74b0\u5883\u3001\u8abf\u9069\u53e5\u6578\u5c0d\u8fa8\u8b58\u7387\u7684\u5f71\u97ff\u4ee5\u63d0\u4f9b\u7d66\u4f7f\u7528\u8005\u5728\u9078\u64c7\u8072\u5b78\u6a21\u578b\u3001\u8a13\u7df4\u3001\u8abf \u9069\u4e0a\u80fd\u5920\u6709\u4e00\u500b\u4f9d\u64da\uff0c\u5229\u7528\u7db2\u969b\u7db2\u8def\u7d50\u5408\u81ea\u52d5\u8a9e\u97f3\u8fa8\u8b58 (Automatic Speech Recognition, ASR) \u7cfb\u7d71\uff0c\u91cb\u51fa\u4e00\u500b\u7db2\u8def\u8a9e\u97f3\u8fa8\u8b58\u7cfb\u7d71\u3002 \u4e8c \u4e8c \u4e8c\u3001 \u3001 \u3001\u7cfb \u7cfb \u7cfb\u7d71 \u7d71 \u7d71\u67b6 \u67b6 \u67b6\u69cb \u69cb \u69cb HMM)\u5982 [8] [9] [10]\uff0c\u8207\u9ad8\u65af\u6df7\u5408\u6a21\u578b (Gaussian Mixture Model, GMM) [11] [12] \u7d71 \u8a08\u6a21\u578b\u7684\u65b9\u6cd5\u5efa\u7acb\u7684\uff0c\u9019\u4e00\u985e\u7684\u8fa8\u8b58\u5de5\u5177\u6709 HTK [13]\u3001CMU Sphinx \u7b49\u7b49\uff0c\u5728\u672c\u7bc7\u8ad6 \u6587\u4e2d\u9078\u64c7\u4f7f\u7528 CMU Sphinx \u7684 Sphinx-4 [14] \u4f5c\u70ba\u6838\u5fc3\u8fa8\u8b58\u5de5\u5177\u3002 of the Twenty-Fifth Conference on Computational Linguistics and Speech Processing (ROCLING 2013) (\u4e00 \u4e00 \u4e00)\u3001 \u3001 \u3001\u8a9e \u8a9e \u8a9e\u6599 \u6599 \u6599\u5eab \u5eab \u5eab\u4ecb \u4ecb \u4ecb\u7d39 \u7d39 \u7d39 \u672c\u7bc7\u8ad6\u6587\u4f7f\u7528 AURORA2 [15] \u7522\u751f\u57fa\u790e\u8072\u5b78\u6a21\u578b (baseline model)\uff0c\u53f0\u7063\u53e3\u97f3\u82f1\u8a9e \u8a9e\u6599 (English Across Taiwan, EAT) [16] \u82f1\u6587\u6578\u5b57\u90e8\u5206 (\u7c21\u7a31 EAT DIGIT) \u53ca\u81ea\u884c\u9304\u88fd \u7684 NOISE1 \u8a9e\u6599\u505a\u70ba\u8abf\u9069\u8a9e\u6599\u9032\u884c\u4e00\u7cfb\u5217\u8abf\u9069\u7684\u5be6\u9a57\u3002", "num": null, "html": null }, "TABREF1": { "type_str": "table", "content": "
(\u4e8c \u4e8c \u4e8c)\u3001 \u3001 \u3001\u5be6 \u5be6 \u5be6\u9a57 \u9a57 \u9a57\u8a2d \u8a2d \u8a2d\u5b9a \u5b9a \u5b9a
\u672c\u5be6\u9a57\u6240\u4f7f\u7528\u7684\u7279\u5fb5\u53c3\u6578\u70ba\u6885\u723e\u5012\u983b\u8b5c\u4fc2\u6578 (Mel-scale Frequency Cepstral Coeffi-
cients, MFCC) \u8868 3\u3001 \u64f7\u53d6\u7279\u5fb5\u53c3\u6578\u6240\u4f7f\u7528\u7684\u53c3\u6578\u6a94\u6848
\u53c3 \u53c3 \u53c3\u6578 \u6578 \u6578\u8aaa \u8aaa \u8aaa\u660e \u660e \u660e\u8a2d \u8a2d \u8a2d\u5b9a \u5b9a \u5b9a\u503c \u503c \u503c
alpha\u9810\u5f37\u8abf\u53c3\u65780.97
dither\u589e\u52a0 1/2-bit \u96dc\u8a0a\u907f\u514d\u96f6\u80fd\u91cf\u97f3\u6846yes
ncep\u5012\u8b5c\u4fc2\u657813
lowerf\u4e0b\u622a\u6b62\u983b\u738764
upperf\u4e0a\u622a\u6b62\u983b\u73874000
nfft\u5feb\u901f\u5085\u5229\u8449\u8f49\u63db\u5927\u5c0f512
wlen\u6f22\u660e\u7a97\u9577\u5ea60.025
input endian \u8f38\u5165\u8cc7\u6599\u7684\u4f4d\u5143\u7d44\u5e8f\uff0c\u5728 NIST \u8207 MS Wav \u683c\u5f0f\u4e2d\u5ffd\u7565big
samprate\u53d6\u6a23\u983b\u73878000
featSphinx \u53c3\u6578\u683c\u5f0f1s c d dd
a) \u5404\u74b0\u5883\u8207\u88dd\u7f6e\u53e5\u6578(b) f5017b \u74b0\u5883\u8a9e\u6599\u5404\u8a9e\u8005\u8207\u88dd\u7f6e\u53e5\u6578
\u5206\u985e\u74b0\u5883\u8a9e\u8005\u4f7f\u7528\u88dd\u7f6e\u8a9e\u8005\u6027\u5225\u4f7f\u7528\u88dd\u7f6e
DHD WFSDHD WFS
cleandormitorytpyen1001Xjcdeng\u59735050
fantpyen50Xmkwu\u75375050
noiseroad basketballtpyen tpyen50 50X Xtpyen yhhuang\u7537 \u753750 5050 50
parkingLottpyen50X
f5017b laboratory \u5982\u8868 2(b) 200200
", "text": "\uff0c\u5982\u8868 3 \u6240\u793a\uff0c\u5728\u64f7\u53d6\u6240\u6709\u97f3\u6a94\u7279\u5fb5\u53c3\u6578\u9664\u4e86\u65bc EAT \u4e2d mic16k \u6240\u4f7f\u7528 \u7684\u53d6\u6a23\u983b\u7387\u70ba 16000 \u5916\u5176\u9918\u6240\u4f7f\u7528\u7684\u53c3\u6578\u90fd\u662f\u4e00\u81f4\u7684\u3002\u5f8c\u7aef\u7684\u8072\u5b78\u6a21\u578b\u4e0a\u4f7f\u7528\u9ad8\u65af \u6df7\u5408\u6a21\u578b\u4f86\u8a13\u7df4\uff0c\u6240\u4f7f\u7528\u7684\u5b57\u5178 (dictionary) \u8207\u97f3\u7d20 (phone) \u5217\u65bc\u8868 4\u4e2d\uff0c\u96b1\u85cf\u5f0f\u99ac\u53ef \u592b\u6a21\u578b\u6bcf\u4e00\u500b\u97f3\u7d20\u4e00\u500b\u6a21\u578b\u3001\u6bcf\u4e00\u500b\u6a21\u578b 3 \u500b\u72c0\u614b\u3001\u6bcf\u500b\u72c0\u614b\u5305\u542b 8 \u500b\u9ad8\u65af\u6df7\u5408\u5206 \u4f48 (Gaussian mixture distribution) \u8a13\u7df4\u4e0a\u4e0b\u6587\u76f8\u95dc (context dependent) \u7684\u6a21\u578b\u3002", "num": null, "html": null }, "TABREF2": { "type_str": "table", "content": "
\u8868 6\u3001 \u5c07 EAT \u516d\u7a2e\u689d\u4ef6\u7684\u8a9e\u6599\u9032\u4e00\u6b65\u5206\u6210\u6e2c\u8a66\u8a9e\u6599\u53ca\u8abf\u9069\u8a9e\u6599 \u8868 9\u3001 AURORA2 \u591a\u74b0\u5883\u8a9e\u6599\u6a21\u578b\u5206\u5225\u4ee5 EAT \u516d\u7a2e\u689d\u4ef6\u505a\u8abf\u9069\u7684\u53e5\u6578\u8207\u6b63\u78ba\u7387 \u8868 7\u3001 \u82f1\u8a9e\u7cfb\u8207\u975e\u82f1\u8a9e\u7cfb\u8154\u8abf\u6bd4\u8f03 \u8868 4\u3001 \u5b57\u5178\u8207\u97f3\u7d20 \u5b57\u5178 \u6587\u5b57 \u97f3\u7d20 \u6e2c\u8a66\u8a9e\u6599 \u8abf\u9069\u8a9e\u6599 \u5973 \u7537 \u7e3d\u53e5\u6578 \u5973 \u7537 \u7e3d\u53e5\u6578 \u7e3d\u8a08 gsmN 106 167 273 106 167 273 546 \u6e2c\u8a66\u8a9e\u6599 \u8abf\u9069\u8a9e\u6599 \u8abf\u9069\u524d 1 5 10 25 50 75 100 \u5168\u90e8 \u566a\u97f3\u74b0\u5883\uff3c\u8072\u5b78\u6a21\u578b clean clean adapt multi multi adapt \u4e7e\u6de8\u8a9e\u6599\u6a21\u578b \u6e2c\u8a66\u8a9e\u6599\uff3c\u8abf\u9069\u8a9e\u6599 gsmN t gsmN a 75.1 75.1 79.9 81.6 84.7 85.7 86.3 86.7 88.1 gsmE t 92.2 91.6 92.0 91.3 92.7 91.6 basketball 65.6 94.6 76.3 96.8 \u4e7e\u6de8\u8a9e\u6599\u6a21\u578b(\u7121\u8abf\u9069) gsmE a gsmNE a gsmE t gsmE a 85.8 84.3 86.8 88.7 89.2 91.2 91.4 91.8 92.2 gsmN t 86.5 88.1 87.4 88.8 86.2 87.3 road 43.0 72.0 64.5 91.491.9 87.4
eight EY eight, T eight gsmE 193 78 271 gsmE test 85.6 pstnN a 78.1 77.6 84.0 86.4 86.0 88.7 88.1 88.5 88.3 193 78 271 542 93.0 92.5 pstnE t pstnN t 90.9 89.6 92.3 91.6 92.4 91.2 fan 62.4 88.2 76.3 97.991.3
five F five, AY five, V five pstnN 84 135 219 gsmN test 73.5 pstnE a 84.1 81.5 87.2 89.3 89.7 90.8 91.5 91.8 92.3 84 136 220 439 86.4 89.5 pstnN t pstnE t 86.4 86.4 88.1 88.3 88.6 88.4 parkingLot 65.6 95.7 66.7 100.0 DHD \u88dd\u7f6e87.7
four F four, OW four, R four pstnE 170 68 238 171 68 \u6e2c\u8a66\u8a9e\u6599\uff3c\u8abf\u9069\u8a9e\u6599 \u4e7e\u6de8\u8a9e\u6599\u6a21\u578b(\u7121\u8abf\u9069) mic16kN t mic16kN a 74.4 75.9 79.8 81.2 84.1 84.7 86.4 87.0 89.5 239 477 pstnE a pstnN a mic16kE t 89.1 88.9 91.0 90.0 92.0 91.6 Avg. 59.2 87.1 71.0 96.5 \u8a9e\u8005\uff3c\u8072\u5b78\u6a21\u578b clean clean adapt multi multi adapt90.4
nine N nine, AY nine, N nine 2 mic16kN 210 348 558 211 349 pstnE test 85.2 mic16kE t mic16kE a 85.1 84.2 86.7 88.0 88.8 90.1 90.8 90.9 92.0 560 1118 92.5 92.7 mic16kN t 83.3 86.4 86.3 87.3 87.4 89.5 cjdeng 54.8 97.9 54.8 97.986.7
oh mic16kE 397 159 OW oh pstnN test Avg. 88.1 \u5c0f\uff0c\u5728\u8f03\u5c0f\u566a\u97f3\u8abf\u9069\u4e0b\u80fd\u5920\u6b63\u5e38\u7684\u5c0d\u8a9e\u8005\u7684\u8154\u8abf\u53e3\u97f3\u53ca\u74b0\u5883\u8abf\u9069\uff0c\u5728\u8f03\u5927\u7684\u566a\u97f3\u4e0b\u5c31 556 397 159 556 1112 77.4 90.5 90.5 88.5 89.5 89.6 89.9 89.9 mkwu 51.6 95.7 50.5 96.889.2
one W one, AX one, N one \u6e2c\u8a66\u8a9e\u6599\uff3c\u8abf\u9069\u8a9e\u6599 \u4e7e\u6de8\u8a9e\u6599\u6a21\u578b(\u7121\u8abf\u9069) \u96fb\u8a71\u76f4\u63a5\u9304\u97f3\u9084\u662f\u900f\u904e\u97f3\u6548\u5361\u4f7f\u7528\u9ea5\u514b\u98a8\u5728\u500b\u4eba\u96fb\u8166\u4e0a\u9304\u97f3\uff0c\u5728\u6c92\u6709\u5176\u5b83\u7279\u5225\u566a\u97f3\u7684 mic16kE a mic16kN a \u6703\u5b8c\u5168\u88ab\u566a\u97f3\u5f71\u97ff\u8b93\u8f49\u79fb\u6a5f\u7387\u7522\u751f\u8f03\u5927\u5e45\u7684\u8b8a\u52d5\uff0c\u4f46\u4f7f\u7528\u591a\u74b0\u5883\u8a9e\u6599\u6a21\u578b\u7684\u9019\u7a2e\u60c5\u6cc1 tpyen 53.8 96.8 54.8 97.6
seven S seven, EH seven, V seven, E seven, N seven six \u8a9e\u6599\u6e2c\u8a66\uff0c\u5f97\u5230\u7684\u82f1\u8a9e\u7cfb\u8207\u975e\u82f1\u8a9e\u7cfb\u8154\u8abf\u5dee\u7570\u505a\u6bd4\u8f03\u5982\u8868 7\uff1a \u7531\u9019\u4e9b\u6578\u64da\u4e2d\u767c\u73fe\u4e0d\u8ad6\u662f\u4f7f\u7528\u82f1\u8a9e\u7cfb\u6216\u975e\u82f1\u8a9e\u7cfb\u7684\u8a9e\u6599\u4f86\u8abf\u9069\uff0c\u8fa8\u8b58\u7387\u90fd\u662f\u82f1\u8a9e\u7cfb mic16kE t 87.1 91.5 92.7 mic16kN t 75.5 87.6 91.2 \u60c5\u6cc1\u4e0b\u4f7f\u7528\u4e0d\u540c\u7684\u53d6\u6a23\u983b\u7387\u5728\u8fa8\u8b58\u7387\u7684\u5dee\u7570\u4e26\u4e0d\u5927\u3002 \u8f03\u4e0d\u660e\u986f\uff0c\u9019\u8a3c\u660e\u4e86\u4e7e\u6de8\u8a9e\u6599\u6a21\u578b\u5728\u4e26\u4e0d\u9069\u5408\u5728\u5c11\u91cf\u4e14\u566a\u97f3\u5927\u7684\u60c5\u6cc1\u4e0b\u9032\u884c\u8abf\u9069\u3002 yhhuang 63.4 100.0 71.0 100.0 \u8868 12\u3001 AURORA2 \u6a21\u578b\u4ee5 NOISE1 clean \u8a9e\u6599\u8abf\u9069\u53e5\u6578\u8207\u6b63\u78ba\u7387 Avg. 55.9 97.6 57.8 98.1 S six, I six, K six, S six 2 three TH three, R three, II three two T two, OO two zero Z zero, II zero, R zero, OW zero filler <s> SIL <sil> </s> \u8a9e\u6599\u512a\u65bc\u975e\u82f1\u8a9e\u7cfb\u8a9e\u6599\u3002 AURORA2 \u7684\u9304\u88fd\u8a9e\u8005\u90fd\u662f\u4ee5\u82f1\u6587\u70ba\u6bcd\u8a9e\uff0c\u56e0\u6b64\u64c1\u6709\u8f03\u6a19\u6e96 \u8868 10\u3001 AURORA2 \u4e7e\u6de8\u8a9e\u6599\u6a21\u578b\u5206\u5225\u4f7f\u7528 EAT \u516d\u7a2e\u689d\u4ef6\u8a9e\u6599\u8abf\u9069\u7684\u8fa8\u8b58\u7387 \u8abf\u9069\u6a21\u578b \u8abf\u9069\u524d 2 5 10 20 25 50 100 150 200 501 \u8868 14\u3001 AURORA2 \u6a21\u578b\u4ee5 NOISE1 noise \u8a9e\u6599\u6e2c\u8a66\u5728\u4e0d\u540c\u566a\u97f3\u74b0\u5883\u8207\u8abf\u9069\u5f8c\u7684\u6b63\u78ba\u7387\u8207 WFS \u88dd\u7f6e \u7684\u53e3\u97f3\u5f97\u5230\u8f03\u9ad8\u7684\u8fa8\u8b58\u7387\u662f\u53ef\u9810\u671f\u7684\u73fe\u8c61\u3002\u5728\u4ee5\u82f1\u8a9e\u7cfb\u8a9e\u6599\u70ba\u6e2c\u8a66\u8a9e\u6599\u7684\u689d\u4ef6\u4e0b\uff0c\u4e0d \u8ad6\u662f\u4f7f\u7528\u82f1\u8a9e\u7cfb\u6216\u975e\u82f1\u8a9e\u7cfb\u8a9e\u6599\u90fd\u80fd\u5f97\u5230\u826f\u597d\u7684\u8abf\u9069\u6548\u679c\uff1b\u76f8\u5c0d\u65bc\u4f7f\u7528\u975e\u82f1\u8a9e\u7cfb\u8a9e\u6599 \u6e2c\u8a66\u6642\u4f7f\u7528\u82f1\u8a9e\u7cfb\u8a9e\u6599\u8abf\u9069\u7684\u6548\u679c\u5c31\u6c92\u90a3\u9ebc\u512a\u7570\u3002 \u9019\u500b\u5be6\u9a57\u7684\u7d50\u679c\u8868\u793a\u51fa\u53e3\u97f3\u5c0d\u8fa8\u8b58\u7387\u7684\u5f71\u97ff\u5f88\u5927\uff0c\u5728\u8a13\u7df4\u904e\u8207\u672a\u8a13\u7df4\u904e\u8fa8\u8b58\u7387\u7684\u5dee \u8ddd\u53ef\u4ee5\u591a\u9054 10%\uff0c\u800c\u4e14\u5373\u4f7f\u662f\u8a13\u7df4\u904e\u7684\u53e3\u97f3\u6216\u591a\u6216\u5c11\u9084\u662f\u6703\u53d7\u5230\u6bcd\u8a9e\u53e3\u97f3\u7684\u5f71\u97ff\uff0c\u5728 \u9078\u64c7\u8abf\u9069\u8a9e\u6599\u7684\u6642\u5f8c\u4ee5\u6311\u9078\u8207\u4f7f\u7528\u8005\u6bcd\u8a9e\u76f8\u540c\u7684\u8a9e\u6599\u70ba\u4f73\u3002 \u591a\u74b0\u5883\u8a9e\u6599\u6a21\u578b \u6e2c\u8a66\u8a9e\u6599\uff3c\u8abf\u9069\u8a9e\u6599 \u591a\u74b0\u5883\u8a9e\u6599\u6a21\u578b(\u7121\u8abf\u9069) gsmE a gsmN a gsmE t 85.8 92.2 91.6 gsmN t 75.1 86.5 88.1 \u6e2c\u8a66\u8a9e\u6599\uff3c\u8abf\u9069\u8a9e\u6599 \u591a\u74b0\u5883\u8a9e\u6599\u6a21\u578b(\u7121\u8abf\u9069) pstnE a pstnN a pstnE t 84.1 92.3 \u4e7e\u6de8\u8a9e\u6599\u6a21\u578b 80.0 81.3 83.6 90.8 93.5 95.1 95.4 97.8 98.3 98.4 98.8 \u8abf\u9069\u53e5\u6578\u95dc\u4fc2 \u8a9e\u8005\uff3c\u8072\u5b78\u6a21\u578b clean clean adapt multi multi adapt \u6e2c\u8a66\u8a9e\u6599\uff3c\u8abf\u9069\u8a9e\u6599 gsmE a gsmN a pstnE a pstnN a mic16kE a mic16kN a Avg. gsmE t 93.0 92.5 92.3 91.4 92.2 91.3 \u591a\u74b0\u5883\u8a9e\u6599\u6a21\u578b 82.5 84.0 88.2 91.1 93.2 95.0 95.1 96.2 96.2 97.3 98.9 \u6a21\u578b clean multi cjdeng 80.6 88.2 75.3 95.7 92.1 gsmN t 86.4 89.5 88.2 87.8 87.1 89.0 \u53e5\u6578 basketball road fan parkingLot Avg. basketball road fan parkingLot Avg. mkwu 78.5 98.9 80.6 100.0 88.0 pstnE t 92.1 92.1 92.5 92.7 92.9 92.7 0 65.6 43.0 62.4 65.6 59.2 76.3 64.5 76.3 66.7 71.0 tpyen 79.6 100.0 83.9 98.9 92.5 pstnN t 88.3 89.8 90.5 90.5 88.7 91.0 ... yhhuang 86.0 100.0 93.5 100.0 89.8 91.6 pstnN t 78.1 88.1 mic16kE t 85.1 92.0 ... 91.6 Avg. 88.9 90.3 90.1 90.2 90.0 91.3 90.1 6 88.2 65.6 81.7 83.9 79.9 84.9 77.4 87.1 82.8 83.1 \u6e2c\u8a66\u8a9e\u6599\uff3c\u8abf\u9069\u8a9e\u6599 \u591a\u74b0\u5883\u8a9e\u6599\u6a21\u578b(\u7121\u8abf\u9069) mic16kE a mic16kN a mic16kN t 83.9 88.0 86.5 87.7 87.6 91.2 87.5 5 87.1 63.4 80.6 83.9 78.8 82.8 76.3 88.2 82.8 82.5 88.3 mic16kE t 89.5 90.0 90.8 90.8 91.5 92.7 90.9 4 82.8 59.1 78.5 74.2 73.7 81.7 75.3 81.7 74.2 78.2 Avg. 81.2 96.8 83.3 98.7
23\u8868 5\u3001 \u4e7e\u6de8\u8a9e\u6599\u8207\u591a\u74b0\u5883\u8a9e\u6599\u6a21\u578b\u8fa8\u8b58\u7387(Avg.\uff1a0-20db \u7684\u5e73\u5747\u503c) mic16kN t 74.4 87.4 89.5 92.5 79.6 92.5 92.5 89.3 95.7 88.2 93.593.592.7
24 2592.5 dB \uff3c\u6e2c\u8a66\u96c6 79.6 92.5 A 94.6 72.0 88.292.5 \u4e7e\u6de8\u8a9e\u6599\u6a21\u578b B C 95.789.3 Avg. 87.6\u591a\u74b0\u5883\u8a9e\u6599\u6a21\u578b 95.7 88.2 93.5 A B C Avg. 96.8 91.4 97.994.6 100.093.0 96.5
clean \u8868 8\u3001 AURORA2 \u4e7e\u6de8\u8a9e\u6599\u6a21\u578b\u5206\u5225\u4ee5 EAT \u516d\u7a2e\u689d\u4ef6\u505a\u8abf\u9069\u7684\u53e5\u6578\u8207\u6b63\u78ba\u7387 99.5 99.5 99.3 99.5 99.0 99.0 98.8 99.0
20 \u6e2c\u8a66\u8a9e\u6599 \u8abf\u9069\u8a9e\u6599 \u8abf\u9069\u524d 95.3 96.8 95.4 95.9 98.3 98.4 97.7 98.2 1 5 10 25 50 75 100 \u5168\u90e8
gsmN t15 gsmE a87.1 90.2 87.1 88.3 97.6 97.5 97.0 97.4 73.5 72.1 80.1 83.8 85.5 87.2 88.4 89.0 89.5
gsmE t10 gsmE a63.6 71.3 64.3 66.8 95.0 95.0 94.3 94.9 85.6 83.6 87.4 89.1 89.1 91.0 91.7 92.6 93.0
pstnN t5 pstnE a24.6 31.6 26.7 27.8 84.5 84.5 84.4 84.5 77.4 76.9 84.5 86.3 87.2 90.2 90.5 90.9 90.5
pstnE t0 pstnE a2.1 85.24.83.5 84.1 87.5 88.4 90.9 92.4 91.9 92.6 92.5 3.5 47.8 48.4 50.5 48.6
-5 mic16kN t mic16kN a0.4 75.51.00.6 77.9 82.4 83.4 85.0 87.1 88.6 89.4 91.2 0.7 8.2 11.7 10.7 10.1
Avg. mic16kE t mic16kE a54.5 58.9 55.4 56.5 84.6 84.8 84.8 84.7 87.1 86.4 87.3 88.6 89.0 91.2 92.2 92.3 91.5
", "text": "Proceedings of the Twenty-Fifth Conference on Computational Linguistics and Speech Processing (ROCLING 2013) 2\u3001 \u3001 \u3001EAT DIGIT \u8abf \u8abf \u8abf\u9069 \u9069 \u9069\u6548 \u6548 \u6548\u679c \u679c \u679c\u8207 \u8207 \u8207\u53e5 \u53e5 \u53e5\u6578 \u6578 \u6578\u95dc \u95dc \u95dc\u4fc2 \u4fc2 \u4fc2 \u5728\u8abf\u9069\u6548\u679c\u8207\u53e5\u6578\u95dc\u4fc2\u7684\u5be6\u9a57\u4e2d\uff0c\u4f7f\u7528\u4e0d\u540c\u7684\u53e5\u6578\u4f86\u89c0\u5bdf\u5404\u500b\u689d\u4ef6\u4e0b\u4f7f\u7528\u4e0d\u540c\u53e5\u6578\u8abf \u9069\u7684\u8fa8\u8b58\u7387\u3002\u5728\u9019\u908a\u8abf\u9069\u53e5\u6578\u55ae\u4f4d 1 \u662f\u4ee3\u8868\u4e00\u53e5\u7537\u751f\u8a9e\u6599\u52a0\u4e0a\u4e00\u53e5\u5973\u751f\u8a9e\u6599 (5 \u55ae\u4f4d\u5c31 \u4ee3\u8868 5 \u53e5\u7537\u751f\u8a9e\u6599\u52a0 5 \u53e5\u5973\u751f\u8a9e\u6599\uff0c\u4ee5\u6b64\u985e\u63a8)\uff0c\u5982\u55ae\u4e00\u6027\u5225\u8a9e\u6599\u4e0d\u8db3\u5247\u4f7f\u7528\u53e6\u4e00\u7a2e\u6027 \u5225\u7684\u8a9e\u6599\u88dc\u8db3\u3002\u5c0d\u4e09\u7a2e\u74b0\u5883\u505a\u53e5\u6578\u8abf\u9069\u6548\u679c\u7684\u6e2c\u8a66\uff0c\u6211\u5011\u8b93\u4e09\u7a2e\u74b0\u5883\u4e2d\u7684\u82f1\u8a9e\u7cfb\u53ca\u975e \u82f1\u8a9e\u7cfb\u7684\u8a9e\u6599\u8f2a\u6d41\u7576\u6e2c\u8a66\u8a9e\u6599\u53ca\u8a13\u7df4\u8a9e\u6599\uff0c\u7d50\u679c\u5982\u8868 8 9\u3002 \u7531\u5be6\u9a57\u7d50\u679c\u4e2d\u80fd\u5920\u770b\u51fa\u8abf\u9069\u53e5\u6578\u8207\u6b63\u78ba\u7387\u6210\u9577\u8d77\u521d\u6b63\u78ba\u7387\u7565\u5fae\u4e0b\u964d\u3001\u5f8c\u4f86\u5feb\u901f \u6210\u9577\u3001\u5230\u6700\u5f8c\u9010\u6f38\u8da8\u7de9\u7684\u6574\u500b\u904e\u7a0b\u8da8\u52e2\uff0c\u7531\u5716\u4e2d\u4e5f\u53ef\u5f97\u77e5\u5728\u4f7f\u7528\u8a9e\u8005\u7121\u95dc (context independent) \u8abf\u9069\u8a9e\u6599\u6642\u4f7f\u7528\u7d04 50 \u55ae\u4f4d (\u7537\u5973\u5404 50 \u53e5) \u7684\u8a9e\u6599\u53ef\u4ee5\u9054\u5230\u6700\u4f73\u7684\u6548\u679c\uff0c\u800c \u4f7f\u7528\u8d85\u904e 50 \u55ae\u4f4d\u5f8c\u8b58\u7387\u7684\u6210\u9577\u4fbf\u9010\u6f38\u8da8\u7de9\uff0c\u5982\u679c\u6211\u5011\u7528\u76f8\u540c\u7684\u65b9\u6cd5\u76f4\u63a5\u4f7f\u7528 50 \u55ae\u4f4d \u7684\u8abf\u9069\u8a9e\u6599\u4f86\u8a13\u7df4\u8072\u6a21\u578b\u53ea\u80fd\u5f97\u5230\u7d04 60% \u7684\u6b63\u78ba\u7387\u3002 3\u3001 \u3001 \u3001EAT DIGIT \u8de8 \u8de8 \u8de8\u74b0 \u74b0 \u74b0\u5883 \u5883 \u5883\u8abf \u8abf \u8abf\u9069 \u9069 \u9069\u6548 \u6548 \u6548\u679c \u679c \u679c \u5728\u9019\u500b\u5be6\u9a57\u4e2d\u6211\u5011\u5c07\u6bcf\u4e00\u7a2e\u689d\u4ef6\u7684\u6e2c\u8a66\u8a9e\u6599\u5206\u5225\u5c0d\u516d\u7a2e\u689d\u4ef6\u7684\u8abf\u9069\u6a21\u578b\u4f86\u6e2c\u8a66\u5728\u9304\u97f3 \u88dd\u7f6e\u8207\u9304\u88fd\u683c\u5f0f\u4e0d\u540c\u7684\u689d\u4ef6\u4e0b\u7684\u5dee\u7570\u6027\uff0c\u6240\u5f97\u5230\u7684\u7d50\u679c\u5982\u8868 10 11\u6240\u793a\uff0cgsm \u8207 pstn \u8a9e \u6599\u90fd\u662f\u85c9\u7531\u96fb\u8a71\u8a71\u7b52\u63a5\u6536\u8072\u97f3\uff0c\u6240\u9304\u5f97\u7684 8KHz 8Bits Mulaw \u683c\u5f0f\u7684\u53d6\u6a23\u9ede\uff0c\u7d93\u7a0b\u5f0f\u8f49 \u62108khz 16bits PCM \u683c\u5f0f\u7684\u53d6\u6a23\u9ede\uff0c\u9ea5\u514b\u98a8\u8a9e\u6599\u5247\u662f\u7531\u500b\u4eba\u96fb\u8166\u53ca\u9ea5\u514b\u98a8\u7d93\u7531\u97f3\u6548\u5361\u9304 \u88fd 16KHz 16bits \u7684\u8072\u97f3\u8a0a\u865f\u3002\u7d50\u679c\u986f\u793a\u9019\u4e9b\u689d\u4ef6\u4e0b\u7684\u74b0\u5883\u662f\u975e\u5e38\u63a5\u8fd1\u7684\uff0c\u4e0d\u8ad6\u662f\u4f7f\u7528 \u5c07\u7db2\u8def\u8fa8\u8b58\u7cfb\u7d71\u904b\u7528\u5728\u73fe\u6d41\u884c\u7684 Android \u624b\u6a5f\u4e0a\u9762\uff0c\u64b0\u5beb\u4e86\u4e00\u500b\u7b26\u5408\u672c\u8ad6\u6587\u4e2d\u6240\u63d0 \u51fa\u7684\u8a9e\u97f3\u8fa8\u8b58\u7cfb\u7d71\u7684\u7528\u6236\u7aef\u7a0b\u5f0f\uff0c\u9032\u884c\u4e00\u7cfb\u5217\u7684\u8fa8\u8b58\u8207\u8abf\u9069\u5be6\u9a57\uff0c\u9019\u4e9b\u5be6\u9a57\u4e3b\u8981\u5728\u6e2c \u8a66\u624b\u6301\u884c\u52d5\u88dd\u7f6e\u4e0a\u4f7f\u7528\u672c\u7bc7\u8ad6\u6587\u4e2d\u7684\u8a9e\u97f3\u8fa8\u8b58\u7cfb\u7d71\u7684\u6548\u80fd\u53ca\u5be6\u7528\u6027\u3002 1\u3001 \u3001 \u3001NOISE1 \u8abf \u8abf \u8abf\u9069 \u9069 \u9069\u6548 \u6548 \u6548\u679c \u679c \u679c\u8207 \u8207 \u8207\u53e5 \u53e5 \u53e5\u6578 \u6578 \u6578 \u9996\u5148\u5c07 NOISE1 \u4e2d\u7684 clean \u8a9e\u6599\u5206\u5169\u500b\u90e8\u5206\uff0c\u5206\u5225\u70ba\u6e2c\u8a66\u8a9e\u6599 (\u524d500\u53e5) \u53ca\u8abf\u9069\u8a9e \u6599 (\u5f8c501\u53e5) \uff0c\u4f7f\u7528\u7684\u8abf\u9069\u8a9e\u6599\u7531\u5c11\u5230\u591a\uff0c\u8abf\u9069\u55ae\u4f4d 1 \u4ee3\u8868\u4e00\u53e5\u8abf\u9069\u8a9e\u6599\uff0c\u5176\u5be6\u9a57\u7d50\u679c \u5982\u8868 12 \u6240\u793a\uff1a \u7531 Android \u88dd\u7f6e\u6240\u9304\u88fd\u7684\u8a9e\u6599\u4e0d\u8ad6\u5728\u4e7e\u6de8\u8a9e\u6599\u6a21\u578b\u6216\u662f\u591a\u74b0\u5883\u8a9e\u6599\u6a21\u578b\u5728\u672a\u8abf\u9069\u7684 \u60c5\u6cc1\u4e0b\u8207\u8a87\u74b0\u5883\u5be6\u9a57\u5f97\u5230\u76f8\u8fd1\u7684\u7d50\u679c\uff0c\u8abf\u9069\u904e\u7a0b\u4e2d\u6240\u4f7f\u7528\u7684\u90fd\u662f\u4f7f\u7528\u540c\u4e00\u500b\u4eba\u7684\u8a9e\u6599 \u4f86\u9032\u884c\uff0c\u5728\u53e5\u6578\u76f8\u540c\u7684\u60c5\u6cc1\u4e0b\u660e\u986f\u5730\u52dd\u904e\u5148\u524d\u4f7f\u7528\u4e0d\u540c\u8a9e\u8005\u8a9e\u6599\u6240\u8abf\u9069\u7684\u6a21\u578b\uff0c\u53e6\u5916 \u7531\u6b64\u8868\u4e2d\u80fd\u89c0\u5bdf\u5230\u7d04\u5728 25 \u5230 50 \u53e5\u6642\u8abf\u9069\u6548\u679c\u9010\u6f38\u8da8\u7de9\uff0c\u56e0\u6b64\u5047\u8a2d\u4ee5 AURORA2 \u7684\u6a21 \u8868 11\u3001 AURORA2 \u591a\u74b0\u5883\u8a9e\u6599\u6a21\u578b\u5206\u5225\u4f7f\u7528 EAT \u516d\u7a2e\u689d\u4ef6\u8a9e\u6599\u8abf\u9069\u7684\u8fa8\u8b58\u7387 \u6e2c\u8a66\u8a9e\u6599\uff3c\u8abf\u9069\u8a9e\u6599 gsmE a gsmN a pstnE a pstnN a mic16kE a mic16kN a Avg. \u578b\u4f7f\u7528 NOISE1 \u8a9e\u6599\u8abf\u9069 25 \u53e5\u80fd\u5f97\u5230\u6700\u5927\u7684\u6295\u8cc7\u5831\u916c\u7387\u7684\u7d50\u679c\u4f86\u9032\u884c NOISE1 \u4e4b\u5f8c\u7684 \u8abf\u9069\u5be6\u9a57\u3002 2\u3001 \u3001 \u3001NOISE1 \u4e0d \u4e0d \u4e0d\u540c \u540c \u540c\u566a \u566a \u566a\u97f3 \u97f3 \u97f3\u74b0 \u74b0 \u74b0\u5883 \u5883 \u5883\u7684 \u7684 \u7684\u8abf \u8abf \u8abf\u9069 \u9069 \u9069\u6548 \u6548 \u6548\u679c \u679c \u679c \u5728\u524d\u9762 NOISE1 \u8207 EAT DIGIT \u8abf\u9069\u5be6\u9a57\u4e2d\u4f7f\u7528\u4e7e\u6de8\u8a9e\u6599\u6a21\u578b\u8207\u591a\u74b0\u5883\u8a9e\u6599\u6a21\u578b\u7684\u5be6 \u9a57\u7d50\u679c\u6c92\u6709\u4ec0\u9ebc\u5dee\u5225\uff0c\u9020\u6210\u9019\u500b\u73fe\u8c61\u7684\u4e3b\u56e0\u5c31\u662f\u56e0\u70ba\u6240\u4f7f\u7528\u7684\u6e2c\u8a66\u8a9e\u6599\u5e7e\u4e4e\u90fd\u662f\u6c92\u6709 \u566a\u97f3\u7684\u8a9e\u6599\uff0c\u800c\u624b\u6301\u5f0f\u88dd\u7f6e\u6700\u65b9\u4fbf\u7684\u4e00\u8655\u5c31\u662f\u8d70\u5230\u54ea\u5c31\u80fd\u5e36\u5230\u54ea\uff0c\u4e0d\u8ad6\u662f\u8981\u5750\u8eca\u3001\u904b \u52d5\u3001\u90ca\u904a\u6216\u662f\u53c3\u52a0\u4e00\u4e9b\u5176\u5b83\u7684\u793e\u4ea4\u6d3b\u52d5\u9019\u4e9b\u88dd\u7f6e\u5e7e\u4e4e\u662f\u5bf8\u6b65\u4e0d\u96e2\u8eab\uff0c\u4f46\u9019\u4e9b\u74b0\u5883\u4e2d \u4e26\u4e0d\u6703\u6bcf\u4e00\u500b\u5730\u65b9\u90fd\u80fd\u8ddf NOISE1 clean \u7684\u74b0\u5883\u4e00\u6a23\u5e7e\u4e4e\u6c92\u6709\u566a\u97f3\uff0c\u53ef\u4ee5\u8aaa\u662f\u6bcf\u4e00\u500b \u74b0\u5883\u4e2d\u90fd\u96e3\u514d\u6703\u6709\u4e00\u4e9b\u566a\u97f3\uff0c\u56b4\u91cd\u7684\u8a71\u751a\u81f3\u807d\u4e0d\u6e05\u695a\u8a9e\u8005\u6240\u8aaa\u7684\u8a71\u3002\u6487\u958b\u9019\u4e9b\u7121\u566a \u97f3\u6216\u566a\u97f3\u6975\u5927\u7684\u6975\u7aef\u60c5\u6cc1\u627e\u5c0b\u751f\u6d3b\u4e0a\u5e38\u5e38\u6703\u9047\u5230\u7684\u5e7e\u7a2e\u566a\u97f3\u4f86\u9032\u884c\u5be6\u9a57\uff0c\u4e00\u5171\u9078\u64c7 \u4e86 basketball\u3001road\u3001fan\u3001parkingLot \u56db\u500b\u74b0\u5883\u566a\u97f3\uff0c\u6bcf\u4e00\u7a2e\u566a\u97f3\u74b0\u5883\u4e0b\u542b\u6709 50 \u53e5\u5747\u4f7f \u7528\u524d 25 \u53e5\u70ba\u6e2c\u8a66\u8cc7\u6599\u5f8c 25 \u53e5\u70ba\u8abf\u9069\u8a9e\u6599\u9032\u884c\u8abf\u9069\u5be6\u9a57\uff0c\u5176\u5be6\u9a57\u7d50\u679c\u5982\u8868 13 \u6240\u793a\uff1a \u5728\u9019\u56db\u7a2e\u74b0\u5883\u4e2d\u53ea\u6709 road \u662f\u5c6c\u65bc\u88ab\u8f03\u5f37\u7684\u566a\u97f3\u6240\u6c61\u67d3\uff0c\u5176\u9918\u4e09\u7a2e\u74b0\u5883\u90fd\u662f\u5c6c\u65bc\u8f15\u5fae \u7684\u566a\u97f3\u5e72\u64fe\u53ef\u4ee5\u5f9e\u4e7e\u6de8\u8a9e\u6599\u6a21\u578b\u7684\u8fa8\u8b58\u7387\u4e2d\u660e\u986f\u7684\u5206\u8fa8\u51fa\u4f86\uff0c\u5373\u4f7f\u662f\u5728\u672a\u91dd\u5c0d\u65b0\u7684\u74b0 \u5883\u4f86\u9032\u884c\u8abf\u9069\u7684\u60c5\u6cc1\u4e0b\u591a\u74b0\u5883\u8a9e\u6599\u6a21\u578b\u4ecd\u7136\u986f\u73fe\u4e86\u4ed6\u5728\u566a\u97f3\u74b0\u5883\u4e0b\u64c1\u6709\u8f03\u597d\u8fa8\u8b58\u7387\u7684 \u512a\u52e2\u3002 \u70ba\u4e86\u9032\u4e00\u6b65\u4e86\u89e3\u5728\u566a\u97f3\u74b0\u5883\u4e4b\u4e0b\u9700\u8981\u591a\u5c11\u8abf\u9069\u8a9e\u6599\u624d\u80fd\u8b93\u9054\u5230\u4e00\u822c\u80fd\u63a5\u53d7\u7684\u6b63\u78ba\u7387 \u9032\u4e00\u6b65\u5c0d\u9019\u4e9b\u8a9e\u6599\u9032\u884c\u53e5\u6578\u8207\u6b63\u78ba\u7387\u7684\u5be6\u9a57\uff0c\u5176\u7d50\u679c\u5982\u8868 14 \u6240\u793a\uff0c\u5c31\u5e73\u5747\u60c5\u6cc1\u4f86\u800c\u8a00 \u91dd\u5c0d\u74b0\u5883\u9032\u884c\u8abf\u9069 5 \u53e5\u4e4b\u5f8c\u80fd\u5920\u5f97\u5230 80% \u5de6\u53f3\u7684\u8fa8\u8b58\u7387\uff0c\u9032\u884c\u5b8c 25 \u53e5\u8abf\u9069\u4e4b\u5f8c\u5c31\u80fd \u5f97\u5230\u7d04 90% \u7684\u6b63\u78ba\u8fa8\u8b58\u7387\u3002\u9019\u5f35\u8868\u683c\u986f\u793a\u4f7f\u7528\u4e7e\u6de8\u8a9e\u6599\u6a21\u578b\u566a\u97f3\u74b0\u5883\u7684\u60c5\u6cc1\u4e0b\u8abf\u9069\u904e \u7a0b\u53cd\u8986\u4e0d\u65b7\u7684\u4e0a\u5347\u4e0b\u964d\uff0c\u9020\u6210\u9019\u7a2e\u60c5\u5f62\u61c9\u8a72\u662f\u56e0\u70ba\u6709\u4e9b\u8abf\u9069\u8a9e\u6599\u566a\u97f3\u8f03\u5927\u800c\u6709\u4e9b\u5247\u8f03 \u8868 13\u3001 AURORA2 \u6a21\u578b\u4ee5 NOISE1 noise \u8a9e\u6599\u6e2c\u8a66\u5728\u4e0d\u540c\u566a\u97f3\u74b0\u5883\u8207\u8abf\u9069\u5f8c\u7684\u6b63\u78ba\u7387 3\u3001 \u3001 \u3001NOISE1 \u4e0d \u4e0d \u4e0d\u540c \u540c \u540c\u88dd \u88dd \u88dd\u7f6e \u7f6e \u7f6e\u7684 \u7684 \u7684\u8abf \u8abf \u8abf\u9069 \u9069 \u9069\u6548 \u6548 \u6548\u679c \u679c \u679c\u6bd4 \u6bd4 \u6bd4\u8f03 \u8f03 \u8f03 \u9664\u4e86\u74b0\u5883\u566a\u97f3\u5c0d\u8fa8\u8b58\u7387\u7684\u5f71\u97ff\u4ee5\u5916\u9084\u8981\u8003\u616e\u5230\u7684\u5c31\u662f\u88dd\u7f6e\u4e0a\u7684\u5dee\u7570\u6027\uff0c\u7562\u7adf\u6bcf\u500b\u88dd\u7f6e \u4e0a\u7684\u9ea5\u514b\u98a8\u54c1\u8cea\u4e0d\u76e1\u76f8\u540c\u3002\u9020\u6210\u8fa8\u8b58\u7387\u5dee\u7570\u7684\u4e0d\u50c5\u50c5\u53ea\u6703\u6709\u9ea5\u514b\u98a8\uff0c\u73fe\u5728\u6709\u4e00\u4e9b\u88dd\u7f6e \u9084\u6703\u81ea\u52d5\u5c07\u8f38\u5165\u97f3\u6e90\u505a\u964d\u566a\u8655\u7406\uff0c\u529f\u80fd\u975e\u5e38\u4eba\u6027\u5316\u4e5f\u975e\u5e38\u7684\u597d\u7528\uff0c\u4f46\u7919\u65bc\u624b\u908a\u6c92\u6709\u9019 \u9ebc\u591a\u88dd\u7f6e\u53ef\u4ee5\u505a\u8fa8\u8b58\u7387\u6e2c\u8a66\u7684\u5be6\u9a57\uff0c\u6211\u5011\u53ea\u53d6\u5f97 DHD \u53ca WFS \u5169\u500b\u88dd\u7f6e\u4f86\u9032\u884c\u5be6\u9a57\uff0c \u4f7f\u7528 NOISE1 \u4e2d\u5206\u985e\u70ba f5017b \u7684\u8a9e\u6599\u6bcf\u4e00\u500b\u8a9e\u8005\u5728\u76f8\u540c\u88dd\u7f6e\u4e4b\u4e0b\u5747\u4f7f\u7528\u524d 25 \u53e5\u70ba\u6e2c \u8a66\u8cc7\u6599\u5f8c 25 \u53e5\u70ba\u8abf\u9069\u8a9e\u6599\u5176\u5be6\u9a57\u7d50\u679c\u5982\u8868 15 \u6240\u793a\u3002\u5f9e\u8868\u4e2d\u6211\u5011\u4e0d\u50c5\u80fd\u89c0\u5bdf\u5230\u82f1\u6587\u767c \u97f3\u9020\u6210\u7684\u5dee\u7570\u4e5f\u80fd\u770b\u5230\u88dd\u7f6e\u4e0d\u540c\u6240\u5e36\u4f86\u7684\u5f71\u97ff\uff0c\u5728\u56db\u4f4d\u8a9e\u8005\u4e2d\u4ee5 yhhuang \u82f1\u6587\u767c\u97f3\u6700 \u70ba\u6a19\u6e96\uff0c\u6240\u5be6\u9a57\u51fa\u4f86\u7684\u8fa8\u8b58\u7387\u679c\u7136\u4e5f\u662f\u6700\u597d\u7684\u3002\u800c\u4e0d\u8ad6\u662f\u4f7f\u7528\u4e7e\u6de8\u8a9e\u6599\u6a21\u578b\u6216\u591a\u74b0\u5883 \u8a9e\u6599\u6a21\u578b\u4ee5 DHD \u88dd\u7f6e\u7684\u8a9e\u6599\u5728\u8abf\u9069\u524d\u7684\u8fa8\u8b58\u7387\u660e\u986f\u4f4e\u65bc WFS \u88dd\u7f6e\uff0c\u5373\u4f7f\u5982\u6b64\u5728\u7d93 \u904e 25 \u53e5\u7684\u74b0\u5883\u8abf\u9069\u4ee5\u5f8c\u5c31\u80fd\u9054\u5230\u5e73\u5747 95% \u4ee5\u4e0a\u7684\u8fa8\u8b58\u7387\uff0c\u85c9\u7531\u9019\u500b\u5be6\u9a57\u6211\u5011\u53ef\u4ee5\u4e86 \u89e3\u5230\u4f7f\u7528\u73fe\u6210\u7684\u8072\u5b78\u6a21\u578b\u65bc\u8154\u8abf\u3001\u4f7f\u7528\u88dd\u7f6e\u4e0d\u540c\u7684\u60c5\u6cc1\u4e5f\u4e0d\u9700\u8981\u7d93\u904e\u5927\u91cf\u7684\u8abf\u9069\u5c31\u80fd \u9054\u5230\u826f\u597d\u7684\u8fa8\u8b58\u7387\u3002 \u8868 15\u3001 AURORA2 \u6a21\u578b\u4ee5 NOISE1 f5017b \u4f7f\u7528\u4e0d\u540c\u88dd\u7f6e\u8a9e\u6599\u8abf\u9069\u53e5\u6578\u8207\u6b63\u78ba\u7387 \u4e94 \u4e94 \u4e94\u3001 \u3001 \u3001\u7d50 \u7d50 \u7d50\u8ad6 \u8ad6 \u8ad6\u8207 \u8207 \u8207\u672a \u672a \u672a\u4f86 \u4f86 \u4f86\u5c55 \u5c55 \u5c55\u671b \u671b \u671b \u672c\u8ad6\u6587\u5229\u7528\u73fe\u6709\u7684\u8a9e\u97f3\u8fa8\u8b58\u5de5\u5177 Sphinx-4 \u6574\u5408\u51fa\u4e00\u500b\u7db2\u8def\u8a9e\u97f3\u8fa8\u8b58\u670d\u52d9\u7cfb\u7d71\uff0c\u9019\u500b \u7cfb\u7d71\u900f\u904e\u7db2\u8def\u63d0\u4f9b\u4e86\u82f1\u6587\u6578\u5b57\u8a9e\u97f3\u8fa8\u8b58\u7684\u670d\u52d9\u4e26\u652f\u63f4\u5feb\u901f\u500b\u4eba\u5316\u529f\u80fd\uff0c\u53ef\u4ee5\u5728\u4e0d\u540c\u74b0 \u5883\u4e2d\u5feb\u901f\u7684\u9054\u5230\u7406\u60f3\u7684\u8fa8\u8b58\u7387\uff0c\u7cfb\u7d71\u5167\u6240\u4f7f\u7528\u7684\u6838\u5fc3\u8fa8\u8b58\u6838\u5fc3 Sphinx-4 \u662f\u7531 JAVA \u8a9e \u8a00\u7de8\u5beb\u800c\u6210\u7684\uff0c\u64c1\u6709\u6975\u5177\u5ef6\u5c55\u6027\u3001\u6a21\u7d44\u5316\u3001\u53ef\u63d2\u62d4\u7684\u67b6\u69cb\u4e26\u4e14\u6709\u826f\u597d\u8de8\u5e73\u53f0\u80fd\u529b\u7684\u512a \u9ede\uff0c\u672c\u8eab\u4e5f\u63d0\u4f9b\u4e86\u8a31\u591a\u7684\u61c9\u7528\u7a0b\u5f0f\u4ecb\u9762\uff0c\u53ef\u4ee5\u8ffd\u8e64\u89e3\u78bc\u5668\u3001\u904b\u884c\u901f\u5ea6\u3001\u8a18\u61b6\u9ad4\u4f7f\u7528\u91cf \u7b49\u7b49\uff0c\u975e\u5e38\u9069\u5408\u7528\u65bc\u7814\u7a76\u3002\u56e0\u70ba Sphinx-4 \u7684\u7279\u6027\u4f7f\u4f3a\u670d\u7aef\u53ef\u4ee5\u5728\u4efb\u4f55\u652f\u63f4 JAVA \u7684\u4f5c \u696d\u7cfb\u7d71\u4e0a\u904b\u884c\uff0c\u800c\u7528\u6236\u7aef\u53ef\u4ee5\u662f\u96fb\u8166\u3001\u624b\u6a5f\u6216\u5176\u5b83\u53ef\u4e0a\u7db2\u7684\u88dd\u7f6e\u3002 \u6b64\u7cfb\u7d71\u900f\u904e\u7db2\u8def\u63d0\u4f9b\u5373\u6642\u7684\u8a9e\u97f3\u8fa8\u8b58\uff0c\u4e26\u4e14\u53ef\u4ee5\u5c07\u4f7f\u7528\u8005\u53ca\u7814\u7a76\u4eba\u54e1\u5c07\u4f7f\u7528\u671f\u9593\u6240 \u8fa8\u8b58\u904e\u7684\u8a9e\u6599\u6536\u96c6\u8d77\u4f86\uff0c\u4f7f\u7528\u4e0a\u975e\u5e38\u5bb9\u6613\u4e14\u65b9\u4fbf\uff0c\u518d\u900f\u5404\u7a2e\u8a9e\u6599\u7684\u8abf\u9069\u5be6\u9a57\u8b93\u4f7f\u7528\u8005 \u5728\u6311\u9078\u8a9e\u8a00\u6a21\u578b\u3001\u8a13\u7df4\u8a9e\u6599\u53ca\u6e2c\u8a66\u8a9e\u6599\u6642\u6709\u500b\u4f9d\u64da\u3002\u5c0d\u65bc\u9019\u500b\u5e73\u53f0\u6211\u5011\u8de8\u51fa\u7684\u7b2c\u4e00\u6b65 \u662f\u5c07\u9019\u500b\u7cfb\u7d71\u6574\u5408\u51fa\u4f86\uff0c\u63d0\u4f9b\u539f\u59cb\u78bc\u8b93\u4efb\u4f55\u6709\u8208\u8da3\u7684\u4eba\u4f7f\u7528\u3002 \u9019\u500b\u7cfb\u7d71\u64c1\u6709\u7db2\u8def\u8a9e\u97f3\u8fa8\u8b58\u3001\u8abf\u9069\u53ca\u8a9e\u6599\u6536\u96c6\u7684\u529f\u80fd\uff0c\u4e26\u80fd\u5920\u5728\u4f7f\u7528\u7684\u904e\u7a0b\u4e2d\u5c07\u8a9e \u6599\u6536\u96c6\u81f3\u4f3a\u670d\u7aef\u3002\u900f\u904e\u7db2\u8def\u8a9e\u97f3\u8fa8\u8b58\u7684\u529f\u80fd\u82e5\u80fd\u52a0\u4e0a\u5176\u5b83\u7684\u6280\u8853\u5c31\u80fd\u884d\u751f\u65b0\u7684\u61c9\u7528\u3002 \u5982\u52a0\u5165\u4eba\u5de5\u667a\u6167\u61c9\u7528\u5728\u667a\u6167\u578b\u624b\u6a5f\u4e0a\uff0c\u5c31\u80fd\u5c55\u73fe\u51fa\u66f4\u5b8c\u5584\u7684\u529f\u80fd\u3002\u800c\u5728\u8a9e\u6599\u6536\u96c6\u9019\u500b \u5340\u584a\u76ee\u524d\u53ea\u662f\u55ae\u7d14\u7684\u628a\u97f3\u6a94\u5132\u5b58\u5728\u4f3a\u670d\u7aef\uff0c\u6c92\u6709\u57f7\u884c\u5206\u985e\u6216\u662f\u904e\u6ffe\u7684\u52d5\u4f5c\uff0c\u5176\u5b83\u529f\u80fd \u4e5f\u9084\u5c1a\u6709\u4e0d\u8db3\u7684\u90e8\u5206\u3002\u4f8b\u5982\u53ef\u4ee5\u5229\u7528\u53ef\u63d2\u62d4\u7684\u7279\u6027\u52a0\u5165\u5c0d\u50b3\u8f38\u6a94\u6848\u9032\u884c\u7de8\u78bc\u58d3\u7e2e\u4f86\u7bc0 \u7701\u7db2\u8def\u983b\u5bec\u3001\u7dda\u4e0a\u5373\u6642\u66f4\u63db\u8072\u5b78\u6a21\u578b\u89e3\u6c7a\u4e0d\u540c\u8a9e\u8a00\u554f\u984c\u3001\u5c0d\u8072\u5b78\u6a21\u578b\u8abf\u9069\u514b\u670d\u4e0d\u540c\u4f7f \u7528\u74b0\u5883\u7b49\u7b49\u529f\u80fd\u3002\u91dd\u5c0d\u4e0a\u8ff0\u5e7e\u9ede\u60c5\u6cc1\u9032\u884c\u64f4\u5145\uff0c\u9019\u500b\u7cfb\u7d71\u5c31\u80fd\u5920\u5438\u5f15\u66f4\u591a\u4eba\u4f7f\u7528\uff0c\u4ee5 \u4fc3\u9032\u8a9e\u97f3\u8fa8\u8b58\u76f8\u95dc\u61c9\u7528\u7814\u7a76\u7684\u767c\u5c55\u3002 \u53c3 \u53c3 \u53c3\u8003 \u8003 \u8003\u6587 \u6587 \u6587\u737b \u737b \u737b", "num": null, "html": null } } } }