| { |
| "paper_id": "2019", |
| "header": { |
| "generated_with": "S2ORC 1.0.0", |
| "date_generated": "2023-01-19T14:54:36.661855Z" |
| }, |
| "title": "Linguistic Analysis for English/Mandarin Speech Synthesis System", |
| "authors": [ |
| { |
| "first": "Yi-Hsiang", |
| "middle": [], |
| "last": "\u6d2a\u7fcc\u7fd4", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "National Pingtung University", |
| "location": {} |
| }, |
| "email": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Hung", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "National Pingtung University", |
| "location": {} |
| }, |
| "email": "" |
| }, |
| { |
| "first": "Yi-Chin", |
| "middle": [], |
| "last": "\u9ec3\u5955\u6b3d", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "National Pingtung University", |
| "location": {} |
| }, |
| "email": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Huang", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "National Pingtung University", |
| "location": {} |
| }, |
| "email": "" |
| }, |
| { |
| "first": "Guang-Feng", |
| "middle": [], |
| "last": "\u9127\u5ee3\u8c50", |
| "suffix": "", |
| "affiliation": {}, |
| "email": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Deng", |
| "suffix": "", |
| "affiliation": {}, |
| "email": "" |
| } |
| ], |
| "year": "", |
| "venue": null, |
| "identifiers": {}, |
| "abstract": "In this study, we analysis the effect of the linguistic information for the English/Mandarin speech synthesis system. In order to construct the acoustic models for both languages, we", |
| "pdf_parse": { |
| "paper_id": "2019", |
| "_pdf_hash": "", |
| "abstract": [ |
| { |
| "text": "In this study, we analysis the effect of the linguistic information for the English/Mandarin speech synthesis system. In order to construct the acoustic models for both languages, we", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Abstract", |
| "sec_num": null |
| } |
| ], |
| "body_text": [ |
| { |
| "text": "adopted the Hidden Markov Model. For the system implementation, we firstly detected the language segments for each language of the input bilingual sentence, and then independently generate the feature sequences for each language. However, for generating fluent synthesized speech, the linguistic information should be taken into account. Here, if the bilingual sentence is mainly written in Mandarin with a few English words, we firstly analyze the Part-Of-Speech information for the English words. Then, we adopted some substitute words (SW) to translate the English parts into Mandarin which have the same POS tags as their corresponding English words. Finally, The entire sentence consists of only one language and could be analyzed linguistically and keep its context information. Finally, the synthesized speech should be more fluent since the contextual linguistic information is used for choosing the suitable acoustic model sequence. In order to construct the original bilingual speech utterance, the English segment is substituted back to the synthesized speech. Experimental results showed that adding the contextual linguistic information is indeed helpful for generating fluent speech for the bilingual sentences. ", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| } |
| ], |
| "back_matter": [], |
| "bib_entries": { |
| "BIBREF0": { |
| "ref_id": "b0", |
| "title": "Unit selection in a concatenative speech synthesis system using a large speech database", |
| "authors": [ |
| { |
| "first": "A", |
| "middle": [ |
| "J" |
| ], |
| "last": "Hunt", |
| "suffix": "" |
| }, |
| { |
| "first": "A", |
| "middle": [ |
| "W" |
| ], |
| "last": "Black", |
| "suffix": "" |
| } |
| ], |
| "year": 1996, |
| "venue": "Acoustics, Speech and Signal Processing", |
| "volume": "1", |
| "issue": "", |
| "pages": "373--376", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "A. J. Hunt and A. W. Black, \"Unit selection in a concatenative speech synthesis system using a large speech database\", in Acoustics, Speech and Signal Processing (ICASSP), vol.1, pp. 373-376, 1996.", |
| "links": null |
| }, |
| "BIBREF1": { |
| "ref_id": "b1", |
| "title": "Simultaneous modeling of spectrum, pitch and duration in HMM-based speech synthesis", |
| "authors": [ |
| { |
| "first": "Takayoshi", |
| "middle": [], |
| "last": "Yoshimura", |
| "suffix": "" |
| } |
| ], |
| "year": 1999, |
| "venue": "Sixth European Conference on Speech Communication and Technology", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Yoshimura, Takayoshi, et al. \"Simultaneous modeling of spectrum, pitch and duration in HMM-based speech synthesis.\" Sixth European Conference on Speech Communication and Technology. 1999.", |
| "links": null |
| }, |
| "BIBREF3": { |
| "ref_id": "b3", |
| "title": "\u5408\u6210\u55ae\u5143\u8207\u554f\u984c\u96c6\u4e4b\u5b9a\u7fa9\u65bc\u96b1\u85cf\u5f0f\u99ac \u53ef\u592b\u6a21\u578b\u4e2d\u6587\u6b4c\u8072\u5408\u6210\u7cfb\u7d71\u4e4b\u5efa\u7acb (Synthesis Unit and Question Set Definition for Mandarin HMM-based Singing Voice Synthesis)", |
| "authors": [ |
| { |
| "first": "Ju-Yun", |
| "middle": [], |
| "last": "Cheng", |
| "suffix": "" |
| }, |
| { |
| "first": "Yi-Chin", |
| "middle": [], |
| "last": "Huang", |
| "suffix": "" |
| }, |
| { |
| "first": "Chung-Hsien", |
| "middle": [], |
| "last": "Wu", |
| "suffix": "" |
| } |
| ], |
| "year": 2013, |
| "venue": "Proceedings of the 25th Conference on Computational Linguistics and Speech Processing", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Ju-Yun Cheng, Yi-Chin Huang, and Chung-Hsien Wu. \"\u5408\u6210\u55ae\u5143\u8207\u554f\u984c\u96c6\u4e4b\u5b9a\u7fa9\u65bc\u96b1\u85cf\u5f0f\u99ac \u53ef\u592b\u6a21\u578b\u4e2d\u6587\u6b4c\u8072\u5408\u6210\u7cfb\u7d71\u4e4b\u5efa\u7acb (Synthesis Unit and Question Set Definition for Mandarin HMM-based Singing Voice Synthesis).\" Proceedings of the 25th Conference on Computational Linguistics and Speech Processing (ROCLING 2013). 2013.", |
| "links": null |
| }, |
| "BIBREF4": { |
| "ref_id": "b4", |
| "title": "\u57fa\u65bc\u96b1\u85cf\u5f0f\u99ac\u53ef\u592b\u6a21\u578b\u4e4b\u4e2d\u6587\u8a9e\u97f3\u5408\u6210\u8207\u543c\u53eb\u60c5\u7dd2\u8f49\u63db", |
| "authors": [], |
| "year": 2010, |
| "venue": "\u6e05\u83ef\u5927\u5b78\u96fb\u6a5f\u5de5\u7a0b\u5b78\u7cfb \u6240\u5b78\u4f4d\u8ad6\u6587", |
| "volume": "", |
| "issue": "", |
| "pages": "1--66", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "\u5433\u5c1a\u9d3b. \"\u57fa\u65bc\u96b1\u85cf\u5f0f\u99ac\u53ef\u592b\u6a21\u578b\u4e4b\u4e2d\u6587\u8a9e\u97f3\u5408\u6210\u8207\u543c\u53eb\u60c5\u7dd2\u8f49\u63db.\" \u6e05\u83ef\u5927\u5b78\u96fb\u6a5f\u5de5\u7a0b\u5b78\u7cfb \u6240\u5b78\u4f4d\u8ad6\u6587 (2010): 1-66.", |
| "links": null |
| }, |
| "BIBREF5": { |
| "ref_id": "b5", |
| "title": "Polyglot speech synthesis based on cross-lingual frame selection using auditory and articulatory features", |
| "authors": [ |
| { |
| "first": "Chia-Ping", |
| "middle": [], |
| "last": "Chen", |
| "suffix": "" |
| }, |
| { |
| "first": "Yi-Chin", |
| "middle": [], |
| "last": "Huang", |
| "suffix": "" |
| }, |
| { |
| "first": "Chung-Hsien", |
| "middle": [], |
| "last": "Wu", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Kuan-De Lee", |
| "suffix": "" |
| } |
| ], |
| "year": 2014, |
| "venue": "Speech, and Language Processing", |
| "volume": "22", |
| "issue": "", |
| "pages": "1558--1570", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Chia-Ping Chen, Yi-Chin Huang, Chung-Hsien Wu, & Kuan-De Lee. (2014). Polyglot speech synthesis based on cross-lingual frame selection using auditory and articulatory features. IEEE/ACM Transactions on Audio, Speech, and Language Processing, 22(10), 1558-1570.", |
| "links": null |
| }, |
| "BIBREF6": { |
| "ref_id": "b6", |
| "title": "DNN-based Bilingual (Telugu-Hindi) Polyglot Speech Synthesis", |
| "authors": [ |
| { |
| "first": "M", |
| "middle": [], |
| "last": "Reddy", |
| "suffix": "" |
| }, |
| { |
| "first": "K. Sreenivasa", |
| "middle": [], |
| "last": "Kiran", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Rao", |
| "suffix": "" |
| } |
| ], |
| "year": 2018, |
| "venue": "International Conference on Advances in Computing, Communications and Informatics (ICACCI)", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Reddy, M. Kiran, and K. Sreenivasa Rao. \"DNN-based Bilingual (Telugu-Hindi) Polyglot Speech Synthesis.\" 2018 International Conference on Advances in Computing, Communications and Informatics (ICACCI). IEEE, 2018.", |
| "links": null |
| }, |
| "BIBREF7": { |
| "ref_id": "b7", |
| "title": "HMM-based mixed-language (Mandarin-English) speech synthesis", |
| "authors": [ |
| { |
| "first": "Y", |
| "middle": [], |
| "last": "Qian", |
| "suffix": "" |
| }, |
| { |
| "first": "H", |
| "middle": [], |
| "last": "Cao", |
| "suffix": "" |
| }, |
| { |
| "first": "F", |
| "middle": [ |
| "K" |
| ], |
| "last": "Soong", |
| "suffix": "" |
| } |
| ], |
| "year": 2008, |
| "venue": "6th International Symposium on Chinese Spoken Language Processing", |
| "volume": "", |
| "issue": "", |
| "pages": "1--4", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Qian, Y., Cao, H., & Soong, F. K. (2008, December). HMM-based mixed-language (Mandarin-English) speech synthesis. In 2008 6th International Symposium on Chinese Spoken Language Processing (pp. 1-4). IEEE.", |
| "links": null |
| }, |
| "BIBREF8": { |
| "ref_id": "b8", |
| "title": "The HMM-based Speech Synthesis System (HTS) Version 2.0", |
| "authors": [ |
| { |
| "first": "H", |
| "middle": [], |
| "last": "Zen", |
| "suffix": "" |
| }, |
| { |
| "first": "T", |
| "middle": [], |
| "last": "Nose", |
| "suffix": "" |
| }, |
| { |
| "first": "J", |
| "middle": [], |
| "last": "Yamagishi", |
| "suffix": "" |
| }, |
| { |
| "first": "S", |
| "middle": [], |
| "last": "Sako", |
| "suffix": "" |
| }, |
| { |
| "first": "K", |
| "middle": [], |
| "last": "Tokuda", |
| "suffix": "" |
| } |
| ], |
| "year": 2007, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Zen, H., Nose, T., Yamagishi, J., Sako, S. and Tokuda, K., The HMM-based Speech Synthesis System (HTS) Version 2.0, 2007. http://hts.sp.nitech.ac.jp/", |
| "links": null |
| }, |
| "BIBREF9": { |
| "ref_id": "b9", |
| "title": "The Hidden Markov Model Toolkit (HTK) Version", |
| "authors": [ |
| { |
| "first": "S", |
| "middle": [], |
| "last": "Young", |
| "suffix": "" |
| }, |
| { |
| "first": "G", |
| "middle": [], |
| "last": "Evermann", |
| "suffix": "" |
| }, |
| { |
| "first": "M", |
| "middle": [], |
| "last": "Gales", |
| "suffix": "" |
| }, |
| { |
| "first": "T", |
| "middle": [], |
| "last": "Hain", |
| "suffix": "" |
| }, |
| { |
| "first": "D", |
| "middle": [], |
| "last": "Kershaw", |
| "suffix": "" |
| }, |
| { |
| "first": "X", |
| "middle": [ |
| "Y" |
| ], |
| "last": "Liu", |
| "suffix": "" |
| }, |
| { |
| "first": "G", |
| "middle": [], |
| "last": "Moore", |
| "suffix": "" |
| }, |
| { |
| "first": "J", |
| "middle": [], |
| "last": "Odell", |
| "suffix": "" |
| }, |
| { |
| "first": "D", |
| "middle": [], |
| "last": "Ollason", |
| "suffix": "" |
| }, |
| { |
| "first": "D", |
| "middle": [], |
| "last": "Povey", |
| "suffix": "" |
| }, |
| { |
| "first": "V", |
| "middle": [], |
| "last": "Valtchev", |
| "suffix": "" |
| }, |
| { |
| "first": "P", |
| "middle": [], |
| "last": "Woodland", |
| "suffix": "" |
| } |
| ], |
| "year": 2006, |
| "venue": "", |
| "volume": "3", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Young, S., Evermann, G., Gales, M., Hain, T., Kershaw, D., Liu, X.Y., Moore, G., Odell, J., Ollason, D., Povey, D., Valtchev, V., and Woodland, P., The Hidden Markov Model Toolkit (HTK) Version 3.4, 2006. http://htk.eng.cam.ac.uk/", |
| "links": null |
| }, |
| "BIBREF10": { |
| "ref_id": "b10", |
| "title": "Reference Manual for Speech Signal Processing Toolkit Ver 3.3", |
| "authors": [ |
| { |
| "first": "", |
| "middle": [], |
| "last": "Sptk Working Group", |
| "suffix": "" |
| } |
| ], |
| "year": null, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "SPTK Working Group, \"Reference Manual for Speech Signal Processing Toolkit Ver 3.3,\". http://sp-tk.sourceforge.net/", |
| "links": null |
| }, |
| "BIBREF11": { |
| "ref_id": "b11", |
| "title": "Phonetic Tutorials", |
| "authors": [ |
| { |
| "first": "T", |
| "middle": [], |
| "last": "Lin", |
| "suffix": "" |
| }, |
| { |
| "first": "L.-J", |
| "middle": [], |
| "last": "Wang", |
| "suffix": "" |
| } |
| ], |
| "year": 1992, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "103--121", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "T. Lin, and L.-J. Wang, \"Phonetic Tutorials\", Beijing University Press, pp. 103-121, 1992.", |
| "links": null |
| }, |
| "BIBREF12": { |
| "ref_id": "b12", |
| "title": "Exploiting prosody hierarchy and dynamic features for pitch modeling and generation in HMM-based speech synthesis", |
| "authors": [ |
| { |
| "first": "C", |
| "middle": [ |
| "C" |
| ], |
| "last": "Hsia", |
| "suffix": "" |
| }, |
| { |
| "first": "C", |
| "middle": [ |
| "H" |
| ], |
| "last": "Wu", |
| "suffix": "" |
| }, |
| { |
| "first": "J", |
| "middle": [ |
| "Y" |
| ], |
| "last": "Wu", |
| "suffix": "" |
| } |
| ], |
| "year": 2010, |
| "venue": "IEEE Transactions on Audio, Speech, and Language Processing", |
| "volume": "18", |
| "issue": "8", |
| "pages": "1994--2003", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Hsia, C. C., Wu, C. H., & Wu, J. Y. (2010). Exploiting prosody hierarchy and dynamic features for pitch modeling and generation in HMM-based speech synthesis. IEEE Transactions on Audio, Speech, and Language Processing, 18(8), 1994-2003.", |
| "links": null |
| }, |
| "BIBREF13": { |
| "ref_id": "b13", |
| "title": "An example of context-dependent label format for HMM-based speech synthesis in English. The HTS CMUARCTIC demo", |
| "authors": [ |
| { |
| "first": "H", |
| "middle": [], |
| "last": "Zen", |
| "suffix": "" |
| } |
| ], |
| "year": 2006, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Zen, H. (2006). An example of context-dependent label format for HMM-based speech synthesis in English. The HTS CMUARCTIC demo, 133.", |
| "links": null |
| }, |
| "BIBREF15": { |
| "ref_id": "b15", |
| "title": "CMU ARCTIC databases for speech synthesis", |
| "authors": [ |
| { |
| "first": "John", |
| "middle": [], |
| "last": "Kominek", |
| "suffix": "" |
| }, |
| { |
| "first": "Alan", |
| "middle": [ |
| "W" |
| ], |
| "last": "Black", |
| "suffix": "" |
| }, |
| { |
| "first": "Ver", |
| "middle": [], |
| "last": "Ver", |
| "suffix": "" |
| } |
| ], |
| "year": 2003, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Kominek, John, Alan W. Black, and Ver Ver. \"CMU ARCTIC databases for speech synthesis.\" (2003).", |
| "links": null |
| } |
| }, |
| "ref_entries": { |
| "TABREF0": { |
| "num": null, |
| "text": "HMM-based Speech Synthesis System, HTS)\uff0c\u662f\u7531 HTS \u5de5\u4f5c\u5718\u968a [9]\u6240\u7814\u767c\uff0c\u8a72\u6280 \u8853\u7531 Hidden Markov Model Toolkit(HTK) [10]\u7814\u767c\u4fee\u6539\u800c\u4f86\uff0cHTS \u5718\u968a\u63d0\u4f9b\u4e86\u4e00\u500b\u4fbf\u65bc", |
| "type_str": "table", |
| "content": "<table><tr><td>\u8fd1\u5e74\u4f86\u8a9e\u97f3\u5408\u6210\u7cfb\u7d71\u5ee3\u6cdb\u88ab\u4f7f\u7528\u7684\u4e3b\u8981\u6709\u5169\u7a2e\uff0c\u4e00\u662f\u57fa\u65bc\u5927\u578b\u8a9e\u6599\u5eab\u6a23\u672c\u505a\u4e32\u63a5\uff0c\u5982: \u6587\u3001\u82f1\u6587\u8a9e\u97f3\u6a21\u578b\u7684\u5b9a\u7fa9\u3002\u7b2c\u4e09\u7ae0 : \u4e2d\u82f1\u593e\u96dc\u8a9e\u97f3\u5408\u6210\u7cfb\u7d71\u5be6\u4f5c\uff0c\u8a73\u8ff0\u4e2d\u82f1\u5207\u5272\u7684\u65b9 \u5176\u4e2d C \u70ba\u7b2c\u4e00\u500b\u97f3\u7d20\u6a21\u578b\u7528\u4f86\u8868\u793a\u8072\u6bcd\u7684\u97f3\uff0cV1+V2 \u5247\u540c\u6642\u7528\u4f86\u8868\u793a\u97fb\u6bcd\u4ee5\u53ca\u4e94\u8072\u8b8a \u505a\u6a19\u8a18\u3002 \u4e09\u4e2d\uff0c\u539f\u53e5\u5b50\u70ba\"\u5bb6\u88e1\u7db2\u8def\u51fa\u4e86\u554f\u984c\u7121\u6cd5\u9023\u4f3a\u670d\u5668\"\u7576\u4e2d\u7684'\u9023'\u5b57\uff0c\u7531 e2 \u53ef\u77e5\u7576\u524d\u97f3\u97fb \u5be6\u9a57\u5c07\u8acb\u53d7\u8a66\u8005\u5206\u5225\u5c0d\u6bcf\u500b\u53e5\u5b50\u4f86\u8a55\u8ad6\u6709\u5206\u6790\u4e26\u8003\u616e\u524d\u5f8c\u6587\u4e14\u8a18\u9304\u6587\u8108\u8a0a\u606f\u4f86\u4e00\u6b21\u5408 \u81f4\u53e5\u5b50\u4e0d\u81ea\u7136\u3002</td></tr><tr><td>\u55ae\u5143\u9078\u64c7(Unit selection approach) [1]\uff0c\u53e6\u4e00\u7a2e\u5247\u662f\u57fa\u65bc\u7d71\u8a08\u65b9\u6cd5\u3002\u5982:\u96b1\u85cf\u5f0f\u99ac\u53ef\u592b\u6a21\u578b \u6cd5\u3001\u4e2d\u6587\u97f3\u7d20\u6a21\u578b\u7684\u5efa\u7acb\u3001\u524d\u5f8c\u6587\u76f8\u95dc\u4e4b\u554f\u984c\u96c6\u6c7a\u7b56\u6a39\u3002\u7b2c\u56db\u7ae0 : \u5be6\u9a57\u7d50\u679c\u53ca\u8a0e\u8ad6\uff0c \u5316\u8da8\u52e2\u7684\u524d(V1)\u5f8c(V2) \uff0c\u4ee5\u6b64\u65b9\u6cd5\u6700\u5f8c\u6211\u5011\u53ef\u80fd\u5982\u4e0b\u5716 107 \u500b\u97f3\u7d20\u6a21\u578b(\u542b\u4e00\u500b pause \u6a21 \u4e09\u3001\u4e2d\u82f1\u593e\u96dc\u8a9e\u97f3\u5408\u6210\u7cfb\u7d71\u5be6\u4f5c \u8a5e\u7531\u56db\u500b\u7d44\u6210\uff0c\u7531 b4 \u53ef\u77e5\u4ed6\u662f\u7531\u524d\u6578\u4f86\u7b2c\u4e00\u500b\u97f3\u7bc0\uff0c\u5373\u4ee3\u8868'\u9023'\u5728\u65b7\u8a5e\u4e2d\u88ab\u65b7\u6210\"\u9023 \u6210\u6574\u500b\u53e5\u5b50\u5f8c\u518d\u66ff\u63db\u6389 SW \u6240\u5408\u51fa\u7684\u8a9e\u97f3\uff0c\u4ee5\u53ca\u8207\u6c92\u6709\u5206\u6790\u524d\u5f8c\u6587\u4e26\u5206\u958b\u5408\u6210\u6bcf\u6bb5\u6587\u5b57 \u5728\u6b64\u4f7f\u7528\u5716\u56db\u4e2d\u6587\u8108\u7684\u53e5\u5b50\"\u5bb6\u88e1\u7db2\u8def\u51fa\u4e86\u554f\u984c\u7121\u6cd5\u9023\u4f3a\u670d\u5668\"\uff0c\u4f46\u5be6\u969b\u4e0a\u5176\u5be6\u6211\u5011</td></tr><tr><td>(HMM-based approach) [2]\u3002\u55ae\u5143\u9078\u64c7\u5408\u6210\u96d6\u7136\u6709\u8457\u6975\u4f73\u7684\u5408\u6210\u97f3\u8cea\uff0c\u4f46\u537b\u9700\u8981\u975e\u5e38\u5927\u91cf \u7684\u8a9e\u6599\u5eab\u505a\u652f\u6301\uff0c\u6240\u4ee5\u5728\u88fd\u4f5c\u8a9e\u6599\u6210\u672c\u4e0a\u6709\u8457\u6975\u5927\u7684\u4ee3\u50f9\u3002\u800c\u96b1\u85cf\u5f0f\u99ac\u53ef\u592b\u6a21\u578b\u5c0d\u8a9e\u6599 \u5eab\u7684\u9700\u6c42\u5247\u4e0d\u50cf\u524d\u8005\u9700\u6c42\u9019\u9ebc\u5927\u3002 \u8a9e\u97f3\u76f8\u95dc\u7684\u61c9\u7528\u4e5f\u975e\u5e38\u591a\uff0c\u4f8b\u5982\u5728\u4e0d\u540c\u8a9e\u901f\u4e0b\u7684\u61c9\u7528 [3]\u3001\u5408\u6210\u6b4c\u5531\u5408\u6210\u7cfb\u7d71 [4]\u3001 \u60c5\u7dd2\u8f49\u63db [5]\u3001\u591a\u8a9e\u8a00\u8a9e\u97f3\u5408\u6210 [6]\u3001\u57fa\u65bc\u6df1\u5ea6\u795e\u7d93\u7db2\u8def\u5728\u591a\u8a9e\u8a00\u9593\u7684\u61c9\u7528 [7]\u7b49\uff0c\u5176\u4e2d \u8aaa\u660e\u5be6\u9a57\u76ee\u7684\u3001\u8a9e\u6599\u5eab\u4f86\u6e90\u4ee5\u53ca\u5167\u5bb9\u3001\u5206\u6790\u53ca\u8a0e\u8ad6\u7d50\u679c\u3002\u7b2c\u4e94\u7ae0 : \u7d50\u8ad6\uff0c\u7e3d\u7d50\u6574\u7bc7\u8ad6 \u6587\u7684\u7d50\u8ad6\u3002 \u578b)\u3002\u76f8\u5c0d\u65bc\u539f\u672c\u5c07\u6240\u6709\u4e2d\u6587\u7d04 1200 \u591a\u500b\u97f3\u7684\u6a21\u578b\u6578\u91cf\uff0c\u6b64\u65b9\u6cd5\u5927\u5927\u6e1b\u5c11\u4e86\u6a21\u578b\u7684\u6578\u91cf \u8868\u4e00\u3001\u97f3\u7d20\u6a21\u578b\u7bc4\u4f8b Syllable C V1 V2 \u310f\u3128\u311f\u3001 huei4 hu eiH erL \u4f3a\u670d\u5668\"\uff1b\u82f1\u6587\u65b9\u9762\u5247\u662f\u5728\u6587\u8108\u5206\u6790\u4e0a\u8207\u8a18\u9304\u97f3\u7bc0\u53ca\u91cd\u97f3(stress)\u65b9\u9762[14]\uff0c\u8207\u4e2d\u6587\u6709\u8457 \u5f8c\u76f4\u63a5\u5408\u4f75\u7684\u8a9e\u97f3\uff0c\u6bd4\u8f03\u54ea\u4e00\u500b\u53e5\u5b50\u8f03\u70ba\u9806\u66a2\u4e26\u8a55\u5206\uff0c\u4f46\u7531\u65bc\u5be6\u9a57\u4e2d\u82f1\u593e\u96dc\u8a9e\u53e5\u6642\u5bb9\u6613 \u60f3\u5408\u7684\u662f Server \u800c\u4e0d\u662f\u4f3a\u670d\u5668\uff0c\u6240\u4ee5\u7576\u4f7f\u7528\u8005\u8f38\u5165\"\u5bb6\u88e1\u7db2\u8def\u51fa\u4e86\u554f\u984c\u7121\u6cd5\u9023 Server\" \u672c\u7cfb\u7d71\u5728\u5408\u6210\u6587\u5b57\u524d\u9808\u5148\u5c07\u539f\u59cb\u6587\u5b57\u505a\u4e2d\u82f1\u6587\u5b57\u5207\u5272\u7684\u524d\u8655\u7406\uff0c\u5728\u5f97\u5230\u4e2d\u82f1\u6587\u5f8c\uff0c\u5148\u5c07 \u985e\u4f3c\u7684\u8a18\u9304\u65b9\u6cd5\u3002 \u4f7f\u53d7\u8a66\u8005\u6df7\u6dc6\uff0c\u6240\u4ee5\u672c\u5be6\u9a57\u50c5\u4ee5\u7d14\u4e2d\u6587\u4e26\u7121\u5c07 SW \u63db\u56de\u82f1\u6587\u7684\u65b9\u6cd5\u4f86\u8b93\u53d7\u8a66\u8005\u63a5\u53d7\u672c\u6b21 \u6642\u5019\uff0c\u6211\u5011\u7684\u6587\u5b57\u524d\u8655\u7406\u5668\u6703\u5148\u5224\u65b7 Server \u662f\u500b\u540d\u8a5e\u4e26\u63db\u6210\u4e00\u500b\u4f5c\u70ba\u540d\u8a5e\u7528\u7684 SW \u5f8c \u6574\u500b\u53e5\u5b50\u4e2d\u7684\u82f1\u6587\u4ee5\u4e00\u500b\u53ef\u8b58\u5225\u7684 SW \u505a\u66ff\u63db\u4e26\u5c07\u6574\u6bb5\u6587\u5b57\u4e1f\u5165\u4e2d\u6587\u5408\u6210\u5668\u76f4\u63a5\u5408\u6210 \u6e2c\u8a66\u3002\u5728\u5be6\u9a57\u4e2d\u6240\u4f7f\u7528\u7684\u8a9e\u6599\u5eab\u5206\u70ba\u4e2d\u82f1\u5169\u90e8\u5206\u3002\u4e2d\u6587\u8a9e\u6599\u5eab\u4f7f\u7528\u8cc7\u7b56\u6703\u6240\u9304\u88fd\u7684\u5973\u6027 \u518d\u5408\u6210\u6574\u53e5\u4e2d\u6587\uff0c\u6b64\u6642\u6587\u8108\u4e2d\u7684'\u9023'\u88ab\u65b7\u6210\"\u9023+\u67d0\u500b\u540d\u8a5e SW\"\uff0c\u5176\u88ab\u7576\u70ba\u52d5\u8a5e\u5f8c\u9762\u6709\u500b \u51fa\u6574\u6bb5\u4e2d\u6587\u53e5\u5b50\u4ee5\u4fdd\u7559\u6574\u6bb5\u4e2d\u6587\u7684\u6587\u8108\u8cc7\u8a0a\u3002\u82f1\u6587\u55ae\u5b57\u5247\u4e1f\u5165\u82f1\u6587\u5408\u6210\u5668\u505a\u5408\u6210\uff0c\u6700\u5f8c \u8a9e\u8005\uff0c\u4e3b\u8981\u5167\u5bb9\u70ba\u65b0\u805e\u8a9e\u6599\uff0c\u7e3d\u5171\u6709 5102 \u500b\u53e5\u5b50\uff0c92388 \u500b\u5b57\uff0c\u82f1\u6587\u8a9e\u6599\u5eab\u65b9\u9762\u4f7f\u7528 \u53d7\u8a5e\u3002\u4f46\u5982\u679c\u6211\u5011\u6c92\u6709\u4f7f\u7528\u6587\u8108\u5206\u6790\uff0c\u5247\u5408\u6210\u53e5\u5b50\u6642\u5019\u6703\u5148\u5408\u6210\"\u5bb6\u88e1\u7db2\u8def\u51fa\u4e86\u554f\u984c\u7121 \u5728\u5c07\u5408\u597d\u7684\u82f1\u6587\u55ae\u5b57\u53d6\u4ee3\u524d\u9762\u7684 SW \u505a\u8a9e\u97f3\u4e32\u63a5\uff0c\u6574\u500b\u67b6\u69cb\u7684\u6d41\u7a0b\u5716\u5982\u5716\u4e8c\u6240\u793a\u3002 (\u4e00)\u3001\u4e2d\u82f1\u593e\u96dc\u6587\u5b57\u5207\u5272 CMU ARCTIC \u8a9e\u6599\u5eab[16]\u3002\u5171\u6709 1132 \u500b\u53e5\u5b50\uff0c\u5176\u4e2d\u5305\u542b 10045 \u500b\u55ae\u5b57(2974 \u500b\u4e0d\u91cd\u8907 \u6cd5\u9023\"\uff0c\u518d\u53e6\u5916\u5408\u6210\u4e00\u500b Server \u5f8c\u76f4\u63a5\u505a\u4e32\u63a5\uff0c\u6b64\u6642\u56e0\u70ba'\u9023'\u70ba\u6700\u5f8c\u4e00\u500b\u5b57\uff0c\u5f8c\u9762\u4e26\u7121</td></tr><tr><td>\u95dc\u65bc [7]DNN \u90e8\u4efd\u6240\u5408\u6210\u7684\u591a\u8a9e\u8a00\u7cfb\u7d71\u70ba\u4f7f\u7528\u55ae\u4e00\u5408\u6210\u5668\uff0c\u4e26\u4e14\u5728\u5408\u6210\u51fa\u4f86\u7684\u8a9e\u97f3\u4e0a\u6709 \u3115\u02ca shr2 shr shrL shrH \u4f7f\u7528\u8005\u8f38\u5165\u6587\u5b57\u5f8c(\u5716\u4e8c. ZH/EN Split \u7684\u6b65\u9a5f)\uff0c\u672c\u7cfb\u7d71\u4e3b\u8981\u4ee5 Unicode \u4f86\u8b58\u5225\u4e2d\u82f1\u6587\u5b57 \u7684\u55ae\u5b57)\uff0c39153 \u500b\u97f3\u7d20\u3002 \u53d7\u8a5e\u4e5f\u6c92\u6709\u5176\u4ed6\u53e5\u5b50\u4f7f\u5176\u6210\u9023\u63a5\u8a5e\uff0c\u6700\u5f8c\u65b7\u8a5e\u5c31\u5c07\u4ed6\u65b7\u6210\"\u7121\u6cd5\u9023\"\uff0c\u5f8c\u9762\u539f\u672c\u8a72\u6709\u7684\u540d</td></tr><tr><td>\u8457\u4e0d\u932f\u7684\u7d50\u679c\uff0c\u4f46\u5176\u7f3a\u9ede\u662f\u5fc5\u9808\u9078\u64c7\u8a9e\u7cfb\u76f8\u4f3c\u7684\u8a9e\u8a00\uff0c\u4e14\u901a\u5e38\u8981\u627e\u5230\u7cbe\u901a\u591a\u8a9e\u8a00\u7684\u76f8\u540c \u3127\u3121\u02c7 yiou3 yi ouL ouL0 \u7684\u5340\u5225\u3002\u5728 Unicode \u4e2d\u4e2d\u6587\u7684\u7bc4\u570d\u70ba 19968~40869\uff0c\u5176\u9918\u90e8\u5206\u5305\u542b\u534a\u5f62\u7a7a\u767d\u7576\u4f5c\u82f1\u6587\u90e8 \u5716\u4e09\u3001Label \u6a94\u7bc4\u4f8b \u672c\u6b21\u5be6\u9a57\u76ee\u7684\u5728\u65bc\u78ba\u8a8d\u6709\u5206\u6790\u524d\u5f8c\u6587\u6587\u8108\u662f\u5426\u6703\u5f71\u97ff\u53e5\u5b50\u7684\u6d41\u66a2\u6027\uff0c\u8a55\u5206\u65b9\u5f0f\u63a1\u5e73 \u8a5e\u88ab\u843d\u55ae\u6210\u4e00\u500b\u55ae\u5b57\u4e86\uff0c\u4f7f\u7684\u6574\u53e5\u9813\u6642\u5931\u53bb\u4e86\u9023\u8cab\uff0c\u5176\u5169\u8005\u983b\u8b5c\u5716\u6bd4\u8f03\u5982\u5716\u516d\uff0c\u5716\u4e2d\u4e0a</td></tr><tr><td>\u8a9e\u8005\u662f\u56f0\u96e3\u7684\uff0c\u5c0d\u6b64\u554f\u984c\u4e5f\u6709 [8]\u9019\u4e9b\u4f7f\u7528\u985e\u4f3c\u97f3\u7d20\u4f86\u9032\u884c\u4e0d\u540c\u8a9e\u8a00\u9593\u8cc7\u6599\u7684\u88dc\u8db3\u7684\u65b9 \u5206(\u975e\u4e2d\u6587)\u3002 2\u3001\u554f\u984c\u96c6 \u5747\u4e3b\u89c0\u503c\u5206\u6578(Mean Opinion Score, MOS)\uff0c\u7531 10 \u4f4d\u53d7\u8a66\u8005\u5c0d 10 \u7d44\u96a8\u6a5f\u9806\u5e8f\u7684\u4e0d\u540c\u7684\u8a9e \u534a\u4e0d\u70ba\u6709\u5206\u6790\u6587\u8108\uff0c\u4e0b\u534a\u90e8\u5247\u7121\u5206\u6790\uff0c\u53ef\u770b\u51fa\u7121\u5206\u6790\u6587\u8108\u6642\uff0c\u9023\u8207 sever \u4e0a\u51fa\u73fe\u4e86\u65b7\u5c64\u3002</td></tr><tr><td>\u6cd5\uff0c\u4e14\u4e2d\u82f1\u5408\u6210\u4e2d\u4e5f\u6709\u4f7f\u7528\u6df7\u5408\u5169\u8a9e\u8a00\u6c7a\u7b56\u6a39\u7684\u65b9\u6cd5\u3002 \u7576\u8f38\u5165\u4e00\u6bb5\u6587\u5b57\u5f8c\uff0c\u9010\u4e00\u7531\u524d\u81f3\u5f8c\u91dd\u5c0d\u6bcf\u4e00\u500b\u6587\u5b57\u505a\u4e2d\u82f1\u6587\u5b57\u5224\u5225\uff0c\u4e26\u4f7f\u7528\u4e00\u500b\u8b8a\u6578\u4f86 \u6839\u64da\u4e2d\u6587\u6a21\u578b\u5b9a\u7fa9\u4ee5\u53ca\u4e0a\u8ff0\u6240\u8a18\u9304\u7684\u6587\u8108\u8cc7\u8a0a\uff0c\u4fbf\u53ef\u958b\u59cb\u8a2d\u8a08\u554f\u984c\u96c6\u4e4b\u6c7a\u7b56\u6a39\u4ee5\u8b93\u6a21 \u53e5\u505a\u8a55\u5206\uff0c\u53d7\u8a66\u8005\u6210\u54e1\u5305\u542b\u540c\u73ed\u540c\u5b78\u3001\u6307\u5c0e\u8001\u5e2b\u3001\u793e\u4ea4\u5e73\u53f0\u4e0a\u7684\u670b\u53cb\u7b49\uff0c\u8a9e\u97f3\u7684\u81ea\u7136\u5ea6 \u6700\u5f8c\u4f7f\u7684\u6574\u53e5\u8a71\u5ff5\u8d77\u4f86\u4e0d\u901a\u9806\uff0c\u7531\u6b64\u53ef\u898b\u6587\u8108\u5206\u6790\u5c0d\u65bc\u4e00\u53e5\u8a71\u7684\u91cd\u8981\u6027\u3002</td></tr><tr><td>\u5132\u5b58\u7576\u524d\u6587\u5b57\u5c6c\u65bc\u4e2d\u6587\u9084\u662f\u82f1\u6587\u7684\u72c0\u614b\u505a\u7d00\u9304\uff0c\u540c\u6642\u6709\u5169\u500b\u9663\u5217\u5206\u5225\u4f86\u5132\u5b58\u7531\u4e0a\u4e00\u6b21\u72c0 \u578b\u9054\u5230\u6700\u4f73\u72c0\u614b\uff0c\u5c0d\u6b64\u6211\u5011\u5c07\u5c0d\u97f3\u7d20\u6a21\u578b\u8003\u616e\u4ee5\u4e0b\u4e94\u5927\u985e\u554f\u984c [15] \u8d8a\u9ad8\u5247\u7d66\u8d8a\u9ad8\u5206(\u6700\u9ad8 5 \u5206)\uff0c\u8d8a\u4e0d\u81ea\u7136\u5247\u7d66\u8d8a\u4f4e\u5206(\u6700\u4f4e 1 \u5206)\uff0c\u6700\u5f8c\u5f97\u5230\u7684 10 \u7d44\u53e5\u5b50\u7684 (\u4e09)\u3001\u96b1\u85cf\u5f0f\u99ac\u53ef\u592b\u6a21\u578b\u7cfb\u7d71\u6982\u8ff0 \u614b\u8f49\u63db\u524d\u81f3\u73fe\u5728\u7684\u4e2d/\u82f1\u6587\u5b57\u6536\u96c6\u5668\uff0c\u7576\u9047\u5230\u72c0\u614b\u8f49\u63db\u6642\u5019\u5c31\u4ee3\u8868\u4e2d\u82f1\u6587\u5b57\u505a\u4e86\u5207\u63db\uff0c (1) \u97f3\u7d20\u76f8\u95dc(Phoneme related): \u82e5\u70ba\u97fb\u6bcd : \u5176\u97f3\u9ad8\u7bc4\u570d(H/M/L)\uff1b\u82e5\u70ba\u8072\u6bcd : \u55ae\u4e00\u6216 \u5e73\u5747\u5206\u6578\u9577\u689d\u5716\u5982\u5716\u56db\u3002\u5176\u4e2d\uff0c\u7d93\u7531\u8a08\u7b97\u5e73\u5747\u7684\u7d50\u679c\u5f8c\u767c\u73fe\uff0c\u6709\u6587\u8108\u5206\u6790\u8f03\u7121\u6587\u8108\u5206\u6790 Keywords: English/Mandarin bilingual sentence, Hidden Markov Model, Linguistic analysis, \u8fd1\u5e74\u4f86\u96b1\u85cf\u5f0f\u99ac\u53ef\u592b\u6a21\u578b\u5728\u8a9e\u97f3\u5408\u6210\u7684\u9818\u57df\u4e2d\u5df2\u7d93\u6709\u8457\u8209\u8db3\u8f15\u91cd\u7684\u5730\u4f4d\uff0c\u56e0\u70ba\u5176\u6240\u5408\u51fa \u5716\u4e00\u3001\u4e2d\u6587\u4e94\u8072\u8b8a\u5316 \u6b64\u6642\u5c31\u628a\u6240\u6709\u6587\u5b57\u6536\u96c6\u5668\u7684\u4e2d/\u82f1\u6587\u4e1f\u5165\u7d50\u679c\u96c6\u5167\uff0c\u7576\u6240\u6709\u6587\u5b57\u90fd\u5224\u5225\u5b8c\u7562\u5f8c\uff0c\u5728\u628a\u6587 \u7531\u5169\u500b\u97f3\u7d20\u7d44\u6210(\u5373\u662f\u5426\u542b\u4ecb\u97f3)\uff1b\u8072\u6bcd\u767c\u97f3\u985e\u5225 : \u585e\u97f3\u3001\u585e\u64e6\u97f3\u3001\u9f3b\u97f3\u3001\u64e6\u97f3\u3001\u908a \u5e73\u5747\u4f86\u5f97\u9ad8(3.94 vs. 3.42) Speech concatenation, Speech synthesis \u4e00\u3001\u7dd2\u8ad6 (\u4e00)\u3001\u7814\u7a76\u52d5\u6a5f \u96a8\u8457\u4e16\u754c\u671d\u5411\u570b\u969b\u5316\u7684\u767c\u5c55\uff0c\u4e0d\u540c\u8a9e\u8a00\u4e4b\u9593\u7684\u4ea4\u6d41\u8d8a\u4f86\u8d8a\u76db\u884c\uff0c\u4e0d\u7ba1\u662f\u5728\u5b78\u754c\u3001\u696d\u754c\uff0c \u90fd\u514d\u4e0d\u4e86\u6703\u63a5\u89f8\u5230\u4e0d\u540c\u8a9e\u8a00\u9593\u7684\u5404\u5f0f\u5404\u6a23\u554f\u984c\uff0c\u800c\u67d0\u4e9b\u7279\u5b9a\u7684\u5c08\u6709\u540d\u8a5e\u7ffb\u6210\u7576\u5730\u8a9e\u8a00\u6642\uff0c \u97f3\u3001\u5507\u97f3\u3001\u820c\u5c16\u97f3\u3001\u820c\u6839\u97f3\u3001\u820c\u9762\u97f3\u3001\u7ff9\u820c\u97f3\u3001\u9f52\u820c\u97f3\uff1b\u97fb\u6bcd\u767c\u97f3\u985e\u5225 : \u55ae\u97fb\u3001 \u7684\u8a9e\u97f3\u6d41\u66a2\u5ea6\u5df2\u7d93\u4e0d\u4e9e\u65bc\u50b3\u7d71\u4e32\u63a5\u5408\u6210\u7684\u7d50\u679c\uff0c\u4e14\u5176\u6240\u8a13\u7df4\u51fa\u4f86\u7684\u6a21\u578b\u6240\u9700\u8981\u7684\u5132\u5b58\u7a7a \u9593\u4e5f\u76f8\u5c0d\u8f03\u5c0f\uff0c\u6709\u8f03\u9ad8\u7684\u651c\u5e36\u6027\u3002\u5728\u672c\u6587\u4e2d\u6240\u4f7f\u7528\u7684\u57fa\u65bc\u96b1\u85cf\u5f0f\u99ac\u53ef\u592b\u6a21\u578b\u7684\u8a9e\u97f3\u5408\u6210 \u5b57\u6536\u96c6\u5668\u5167\u7684\u7d50\u679c\u653e\u5165\u7d50\u679c\u96c6\u5c31\u5c07\u539f\u59cb\u53e5\u5b50\u5207\u5272\u5b8c\u6210\u3002 \u4e8c\u3001\u4e2d\u82f1\u6587\u8a9e\u97f3\u6a21\u578b\u5b9a\u7fa9 \u8907\u97fb\u3001\u8072\u96a8\u97fb\u3001\u6372\u820c\u97fb\uff1b\u97f3\u7d20\u5728\u97f3\u7bc0\u4e2d\u7684\u4f4d\u7f6e : \u7531\u524d/\u5f8c\u6578\u4f86\u4f4d\u7f6e\u3002 MOS \u4e3b\u89c0\u8a55\u6e2c (\u4e8c)\u3001\u8a9e\u97f3\u4e32\u63a5 (\u4e00)\u3001\u4e2d\u6587\u6a21\u578b\u5b9a\u7fa9 (2) \u97f3\u7bc0\u76f8\u95dc(Syllable related): \u97f3\u7bc0\u4e2d\u97f3\u7d20\u7684\u6578\u91cf : \u8003\u616e\u524d/\u7576\u524d/\u4e0b\u4e00\u500b\u97f3\u7bc0\uff1b\u5728\u97f3\u97fb 5 \u7cfb\u7d71(\u958b\u767c\u7684\u7814\u7a76\u5e73\u53f0\uff0c\u4e26\u6709\u6548\u5e6b\u52a9 HMM \u7684\u8a13\u7df4\u3002 \u672c\u7814\u7a76\u76ee\u6a19\u70ba\u5efa\u7acb\u4e00\u5957\u57fa\u65bc HMM \u7684\u8a9e\u97f3\u6a21\u578b\uff0c\u4e3b\u8981\u53ef\u5206\u70ba\u8a13\u7df4\u8207\u5408\u6210\u5169\u500b\u968e\u6bb5\u3002 \u4e2d\u6587\u7d04\u6709 420 \u500b\u4e0d\u542b\u8072\u8abf\u7684\u57fa\u672c\u55ae\u5143\uff0c\u52a0\u5165\u4e94\u8072\u8b8a\u5316\u5f8c\uff0c\u5247\u6709\u8d85\u904e 1200 \u500b\u542b\u8072\u8abf\u7684\u55ae \u5143\u3002\u82e5\u76f4\u63a5\u5c0d\u9019\u7d04 1200 \u500b\u6a21\u578b\u9032\u884c\u8a13\u7df4\u6216\u8a31\u53ef\u4ee5\u9054\u5230\u4e0d\u932f\u7684\u7d50\u679c\uff0c\u4f46\u662f\u9019\u9700\u8981\u975e\u5e38\u975e \u5e38\u5927\u91cf\u7684\u8a9e\u6599\u5eab\u4f86\u8a13\u7df4\u624d\u6709\u53ef\u80fd\uff0c\u82e5\u8cc7\u6599\u91cf\u4e0d\u8db3\u53ef\u80fd\u5c0e\u81f4\u67d0\u4e9b\u97f3\u8a13\u7df4\u4e0d\u8db3\u96e3\u4ee5\u767c\u51fa\u6b63\u78ba \u7684\u97f3\u8abf\uff0c\u751a\u81f3\u53ef\u80fd\u767c\u751f\u67d0\u4e9b\u97f3\u9023\u4e00\u500b\u8cc7\u6599\u90fd\u6c92\u6709\u7684\u60c5\u6cc1\u3002 1\u3001\u6587\u8108\u5206\u6790 4.5 \u8a5e\u3001\u97f3\u97fb\u7247\u8a9e\u4e2d\u7684\u4f4d\u7f6e : \u7531\u7531\u524d/\u5f8c\u6578\u4f86\u4f4d\u7f6e\u3002 4 \u70ba\u4e86\u4f7f\u4e2d\u82f1\u6587\u9023\u63a5\u8655\u6709\u8457\u66f4\u81ea\u7136\u7684\u767c\u97f3\u8207\u97fb\u5f8b\uff0c\u6211\u5011\u5fc5\u9808\u4fdd\u6709\u6574\u53e5\u4e2d\u6587\u5b57\u4e2d\u7684\u524d\u5f8c\u95dc 3.5 (3) \u97f3\u97fb\u8a5e\u76f8\u95dc(Prosodic Word related): \u97f3\u97fb\u8a5e\u4e2d\u97f3\u7bc0\u7684\u6578\u91cf : \u8003\u616e\u524d/\u7576\u524d/\u4e0b\u4e00\u500b\u97f3 3 \u4fc2\u4e26\u8a18\u9304\u4e0b\u4f86\uff0c\u56e0\u6b64\u5728\u5408\u6210\u8a9e\u97f3\u524d\u5fc5\u9808\u5148\u5c0d\u6587\u5b57\u505a\u524d\u8655\u7406(\u5716\u4e8c. ZH/EN Text Analysis)\uff0c \u70ba\u6b64\u6211\u5011\u5fc5\u9808\u5b9a\u7fa9\u4ee5\u4e0b\u4e2d\u6587\u6587\u8108\u8cc7\u8a0a\uff0c\u4e26\u52a0\u4ee5\u8a18\u9304\u6210 Label \u6a94\u3002\u7d30\u7bc0\u5982\u4e0b\u6240\u5217: \u97f3\u7d20 2.5 \u97fb\u8a5e\uff1b\u97f3\u97fb\u8a5e\u5728\u97f3\u97fb\u7247\u8a9e\u4e2d\u7684\u4f4d\u7f6e : \u7531\u524d/\u5f8c\u6578\u4f86\u4f4d\u7f6e\u3002 (4) \u97f3\u97fb\u77ed\u8a9e\u76f8\u95dc(Prosodic Phrase related): \u97f3\u97fb\u7247\u8a9e\u4e2d\u97f3\u7bc0\u3001\u97f3\u97fb\u8a5e\u7684\u6578\u91cf : \u8003\u616e\u524d/ 2 \u5716\u516d\u3001\u6587\u8108\u5206\u6790\u983b\u8b5c\u5716\u6bd4\u8f03 1.5 \u6642\u5e38\u6703\u6709\u7121\u6cd5\u5145\u5be6\u8868\u9054\u5176\u539f\u610f\u7684\u56f0\u5883\uff0c\u4ea6\u6216\u8005\u53ef\u80fd\u767c\u751f\u985e\u4f3c\u7e41\u4e2d\u8207\u7c21\u4e2d\u7684\u7ffb\u8b6f\u5b8c\u5168\u4e0d\u540c \u751a\u81f3\u5230\u4e86\u76f8\u53cd\u7684\u610f\u7fa9(\u5982:'\u884c'\u8207'\u5217')\u3002\u6b64\u6642\u70ba\u4e86\u907f\u514d\u554f\u984c\uff0c\u6211\u5011\u5e38\u5e38\u6703\u5728\u8a9e\u8a00\u4e2d\u76f4\u63a5\u4f7f \u7528\u8a72\u8a5e\u7684\u539f\u672c\u767c\u97f3\u4f86\u6e9d\u901a\uff0c\u9019\u5728\u4eba\u8207\u4eba\u4e4b\u9593\u6216\u8a31\u4e26\u6c92\u6709\u4ec0\u9ebc\u592a\u5927\u7684\u554f\u984c\uff0c\u4f46\u5be6\u969b\u4e0a\u56e0\u70ba \u4e2d\u6587\u8ddf\u82f1\u6587\u4e0d\u7ba1\u5728\u767c\u97f3\u9084\u662f\u6574\u500b\u8a9e\u8a00\u4ee5\u5373\u6587\u5b57\u7684\u7d50\u69cb\u4e0a\u90fd\u6709\u8457\u975e\u5e38\u5927\u7684\u4e0d\u540c\uff0c\u9019\u4f7f\u7684\u5982 \u679c\u8981\u5408\u6210\u4e00\u53e5\u591a\u8a9e\u8a00\u7684\u8a9e\u97f3\u5408\u6210\u5bb9\u6613\u767c\u751f\u8a9e\u8abf\u4e0d\u9806\u66a2\u7684\u554f\u984c\uff0c\u70ba\u6b64\u6211\u5011\u5fc5\u9808\u627e\u4e00\u500b\u65b9\u6cd5 \u4f86\u89e3\u6c7a\u6b64\u554f\u984c\u3002\u672c\u8ad6\u6587\u63d0\u51fa\u4e00\u7a2e\u4f9d\u7167\u524d\u5f8c\u6587\u5206\u6790\u6587\u8108\u95dc\u4fc2\u4ee5\u53ca\u6839\u64da\u5176\u8a5e\u6027\u4f86\u78ba\u8a8d\u4e2d\u6587\u7684 \u767c\u97f3\u65b9\u5f0f\uff0c\u4f7f\u5f97\u5408\u6210\u4e2d\u82f1\u593e\u96dc\u8a9e\u53e5\u6642\u80fd\u4fdd\u6301\u4e2d\u6587\u7684\u6574\u9ad4\u8108\u7d61\uff0c\u9032\u800c\u63d0\u5347\u5408\u6210\u8a9e\u97f3\u7684\u6d41\u66a2 \u6027\u8207\u81ea\u7136\u5ea6\u3002 (\u4e8c)\u3001\u76f8\u95dc\u7814\u7a76 \u8a13\u7df4\u968e\u6bb5\u6642\u7531\u8072\u97f3\u8a9e\u6599\u85c9\u7531 SPTK [11]\u4f30\u7b97\u983b\u8b5c\u53ca\u97f3\u97fb\u53c3\u6578\uff0c\u8072\u97f3\u8a9e\u6599\u6240\u76f8\u5c0d\u61c9\u7684\u6587\u5b57 \u7d93\u7531\u6587\u5b57\u5206\u6790\u5668(Text analysis)\u7522\u751f\u76f8\u61c9\u7684\u6587\u8108\u8a0a\u606f\uff0c\u5728\u7d93\u7531\u554f\u984c\u96c6(Question set)\u5206\u985e\u6a39 \u7522\u751f\u8207\u6587\u8108\u8a0a\u606f\u76f8\u95dc\u7684 HMM \u6a21\u578b\u3002\u5408\u6210\u968e\u6bb5\u5c07\u6b32\u5408\u6210\u7684\u8a9e\u97f3\u6587\u5b57\u7576\u4f5c\u8f38\u5165\u4e1f\u5165\u6587\u5b57\u5206 \u6790\u5668\u4e2d\u7522\u751f\u76f8\u61c9\u7684\u6587\u8108\u8a0a\u606f\uff0c\u518d\u7d93\u7531\u554f\u984c\u96c6\u5206\u985e\u6a39\u627e\u51fa\u8a72\u6bb5\u6587\u5b57\u5c0d\u61c9\u7684 HMM \u6a21\u578b\u5e8f\u5217\uff0c \u518d\u7d93\u7531 HMM \u6a21\u578b\u7522\u751f\u5c0d\u61c9\u7684\u983b\u8b5c\u53ca\u97f3\u97fb\u53c3\u6578\uff0c\u6700\u5f8c\u5408\u6210\u51fa\u6240\u9700\u7684\u8a9e\u97f3\u8a0a\u865f\u7576\u4f5c\u8f38\u51fa\u3002 (\u56db)\u3001\u7ae0\u7bc0\u6982\u8ff0 \u672c\u8ad6\u6587\u4e3b\u8981\u6558\u8ff0\u4e2d\u82f1\u593e\u96dc\u8a9e\u97f3\u5408\u6210\u7cfb\u7d71\u70ba\u4e86\u78ba\u4fdd\u9023\u8cab\u6027\uff0c\u800c\u4f7f\u7528 SW \u66ff\u63db\u82f1\u6587\u5b57\u4f86\u5408\u6210 \u6574\u53e5\u4e2d\u6587\u5b57\u4ee5\u4fdd\u6301\u6574\u9ad4\u4e2d\u6587\u8108\u7d61\uff0c\u672c\u8ad6\u6587\u4e3b\u8981\u5206\u6210\u4e94\u7bc0:\u7b2c\u4e00\u7ae0 : \u7dd2\u8ad6\uff0c\u4e3b\u8981\u8aaa\u660e\u7814\u7a76 \u52d5\u6a5f\u3001\u76f8\u95dc\u7814\u7a76\u8207\u8a0e\u8ad6\u3001\u4ee5\u53ca\u7cfb\u7d71\u6982\u8ff0\u3002\u7b2c\u4e8c\u7ae0 : \u4e2d\u82f1\u6587\u8a9e\u97f3\u6a21\u578b\uff0c\u5b9a\u7fa9\u5206\u5225\u4ecb\u7d39\u4e2d (Extended Final):\u5728\u4e2d\u6587\u4e2d\u5373\u8868\u793a\u4e00\u500b\u8072\u6bcd/\u97fb\u6bcd/\u4ecb\u97f3\uff0c\u5982:\u3105\u3001\u3106\u3001\u3127\u3001\u3128\u3001\u3122\u3001\u3125 \uff1b 1 \u70ba\u4e86\u89e3\u6c7a\u6a21\u578b\u904e\u591a\u5c0e\u81f4\u8a13\u7df4\u4e0d\u8db3\u7684\u554f\u984c\u6211\u5011\u5fc5\u9808\u76e1\u53ef\u80fd\u58d3\u4f4e\u6a21\u578b\u7684\u6578\u91cf\uff0c\u672c\u8ad6\u6587\u4e2d\uff0c \u6211\u5011\u4ee5\"Segmental Tonal Phone Model\" (STPM) [12]\u505a\u70ba\u5b9a\u7fa9\u6211\u5011\u4e2d\u6587\u6a21\u578b\u7684\u65b9\u6cd5\uff0c\u5c0d\u6b64 \u6211\u5011\u5fc5\u9808\u5c07\u6a21\u578b\u8003\u616e\u81f3\u8072\u6bcd\u3001\u97fb\u6bcd\u4ee5\u53ca\u4e94\u8072\u4f86\u9032\u884c\u8a2d\u8a08\u3002\u7136\u800c\u4e94\u8072\u7684\u8b8a\u5316\u5dee\u5225\u5728\u97f3\u9ad8\u4e0a\uff0c \u6240\u4ee5\u8981\u5340\u5225\u4e94\u8072\u5c31\u5fc5\u9808\u5f97\u89c0\u5bdf\u4e94\u8072\u97f3\u9ad8\u7684\u7684\u8b8a\u5316\uff0c\u5716\u4e00\u5206\u5225\u70ba\u4e94\u8072\u7684\u983b\u8b5c\u5716\u4ee5\u53ca\u5176\u97f3\u9ad8 (\u5716\u4e2d\u8072\u97f3\u5206\u5225\u70ba:\u5df4\u3001\u62d4\u3001\u628a\u3001\u7238\u3001\u5427)\u3002\u5982\u5716\u6240\u793a\uff0c\u97f3\u983b\u7684\u7bc4\u570d\u5927\u6982\u53ef\u5206\u70ba\u9ad8\u97f3\u983b\u7bc4\u570d \u5716\u4e8c\u3001\u4e2d\u82f1\u593e\u96dc\u5408\u6210\u7cfb\u7d71\u6d41\u7a0b\u5716 \u7576\u524d/\u4e0b\u4e00\u500b\u97f3\u97fb\u7247\u8a9e\uff1b\u97f3\u97fb\u7247\u8a9e\u5728\u6574\u6bb5\u8a71\u7684\u4f4d\u7f6e : \u7531\u524d/\u5f8c\u6578\u4f86\u4f4d\u7f6e\u3002 0.5 \u4e94\u3001\u7d50\u8ad6 \u97f3\u7bc0(Syllable):\u5728\u4e2d\u6587\u4e2d\u5373\u8868\u793a\u4e00\u500b\u5b57\uff0c\u5982:\u6211\u3001\u525b\u3001\u56de\u3001\u5bb6\uff1b\u97f3\u97fb\u8a5e(Prosodic Word) \u591a\u500b\u97f3\u7bc0\u7d44\u5408\u6210\u7684\u4e00\u500b\u8a5e\uff0c\u5982: \u879e\u87fb\u3001\u7121\u6cd5\u3001\u6211\u5011\uff1b\u97f3\u97fb\u77ed\u8a9e(Prosodic Phrase): \u591a\u500b 0 (5) \u53e5\u5b50\u76f8\u95dc(Utterance related): \u53e5\u5b50\u4e2d\u97f3\u7bc0\u3001\u97f3\u97fb\u8a5e\u3001\u97f3\u97fb\u77ed\u8a9e\u7684\u6578\u91cf\u3002 1 2 3 4 5 6 7 8 9 10 \u672c\u8ad6\u6587\u70ba\u4e2d\u82f1\u593e\u96dc\u8a9e\u97f3\u5408\u6210\u7cfb\u7d71\uff0c\u5728\u4e00\u53e5\u4e2d\u82f1\u6df7\u96dc\u7684\u8a9e\u53e5\u4ee5 SW \u66ff\u63db\u82f1\u6587\u5b57\u4e26\u85c9\u6b64\u5408\u51fa \u7576\u5176\u5728\u8a13\u7df4\u968e\u6bb5\u900f\u904e\u554f\u984c\u96c6\u5206\u985e\u4e26\u8a13\u7df4\u5b8c\u6210\u6a21\u578b\u5f8c\uff0c\u5728\u5408\u6210\u8a9e\u97f3\u63a5\u6bb5\u6642\u5c31\u53ef\u518d\u6b21\u5f9e\u6b64 \u6709\u6587\u8108\u5206\u6790 \u7121\u6587\u8108\u5206\u6790 (\u4e8c)\u3001\u82f1\u6587\u6a21\u578b\u5b9a\u7fa9 \u97f3\u97fb\u8a5e\u7d44\u6210\u7684\u4e00\u5c0f\u6bb5\u8a71\uff0c\u901a\u5e38\u5169\u500b\u77ed\u8a9e\u9593\u6703\u6709\u505c\u9813\uff0c\u6240\u4ee5\u77ed\u8a9e\u5f88\u5bb9\u6613\u767c\u751f\u5728\u9023\u63a5\u8a5e\u4e0a\uff1b \u6574\u53e5\u8003\u616e\u524d\u5f8c\u6587\u95dc\u4fc2\u7684\u4e2d\u6587\u4ee5\u4fdd\u8b49\u53e5\u5b50\u7684\u6d41\u66a2\u6027\uff0c\u6700\u5f8c\u5728\u4ee5\u5408\u6210\u7684\u82f1\u6587\u63db\u6389\u66ff\u63db\u7528\u7684 \u554f\u984c\u96c6(\u5716 3. Question Set)\u4e2d\u5f9e\u6240\u9700\u5408\u6210\u6587\u5b57\u7684 Label \u4e2d\u627e\u5230\u5176\u5c0d\u61c9\u7684\u6a21\u578b\u4e26\u5408\u6210\u8a72\u8a9e \u82f1\u6587\u6a21\u578b\u65b9\u9762\u5247\u76f4\u63a5\u57fa\u65bc ARPAbet \u6a19\u793a\u4e26\u4ee5\u570b\u969b\u97f3\u6a19(IPA)\u70ba\u6e96\u4f86\u505a\u70ba\u6211\u5011\u7684\u97f3\u7d20\u6a21\u578b\uff0c \u53e5\u5b50(Utterance): \u4e00\u500b\u4ee5\u4e0a\u7684\u77ed\u8a9e\u7d44\u6210\u4e00\u500b\u53e5\u5b50\uff0c\u5373\u6211\u5011\u8981\u5408\u6210\u7684\u76ee\u6a19\u8a9e\u53e5\u3002\u4ee5\u4e0a\u6587 \u5716\u4e94\u3001MOS \u4e3b\u89c0\u8a55\u4f30 SW \u4f86\u9084\u539f\u539f\u672c\u7684\u4e2d\u82f1\u593e\u96dc\u53e5\u5b50\u3002\u5be6\u4f5c\u65b9\u6cd5\u4ee5\u96b1\u85cf\u5f0f\u99ac\u5361\u592b\u6a21\u578b\u4e26\u5728\u97f3\u7d20\u6a21\u578b\u4ee5\u53ca\u53e5\u5b50 \u97f3\uff0c\u6700\u5f8c\u5728\u5c07\u4e2d\u82f1\u6587\u90e8\u5206\u5408\u4f75\uff0c\u7531\u82f1\u6587\u5b57\u53d6\u4ee3 SW \u4e26\u5f97\u5230\u6700\u4e2d\u7684\u8a9e\u97f3\u8f38\u51fa\u7d50\u679c\u3002 \u5176\u4e2d\u5305\u542b 13 \u500b\u5143\u97f3\u30013 \u500b\u96d9\u5143\u97f3\u300127 \u500b\u8f14\u97f3\uff0c\u5171 43 \u500b\u97f3\u3002\u5143\u97f3\u53c8\u56e0\u767c\u97f3\u4f4d\u7f6e\u53ef\u7d30\u5206\u81f3 \u8108\u8cc7\u8a0a\u6703\u4f9d\u7167\u5176\u524d\u524d\u4e00\u500b\u55ae\u5143\u3001\u524d\u4e00\u500b\u55ae\u5143\u3001\u7576\u524d\u55ae\u5143\u3001\u4e0b\u4e00\u500b\u55ae\u5143\u3001\u4e0b\u4e0b\u4e00\u500b\u55ae\u5143 \u7684\u6587\u8108\u8cc7\u8a0a\u4f7f\u7528\u554f\u984c\u96c6\u4e4b\u6c7a\u7b56\u6a39\u4f86\u9032\u884c\u6a21\u578b\u512a\u5316\u3002 (H)\u3001\u4e2d\u97f3\u983b\u7bc4\u570d(M)\u3001\u4f4e\u97f3\u983b\u7bc4\u570d(L)\uff0c\u4e94\u8072\u7684\u8b8a\u5316\u8da8\u52e2\u5206\u5225\u70ba: \u4e00\u8072:H\u2192H \u3001 \u4e8c\u8072:L\u2192H \u3001 \u4e09\u8072:L\u2192L \u3001 \u56db\u8072:H\u2192L \u3001 \u8f15\u8072:M\u2192M \u70ba\u4e86\u80fd\u8868\u9054\u4e94\u8072\u7684\u8b8a\u5316\u8da8\u52e2\uff0c\u6211\u5011\u5c07\u4e2d\u6587\u7684\u57fa\u672c\u55ae\u5143\u5206\u5272\u6210\u4e09\u500b\u97f3\u7d20\u6a21\u578b: C+V1+V2 \u524d\u3001\u4e2d\u3001\u5f8c\u5143\u97f3\uff0c\u8f14\u97f3\u4e5f\u53ef\u4f9d\u767c\u97f3\u4f4d\u7f6e\u5206\u6210\u585e\u97f3\u3001\u64e6\u97f3\u3001\u534a\u5143\u97f3\u3001\u9f3b\u97f3\u3001\u585e\u64e6\u97f3\u7b49\u3002 (\u4e8c)\u5206\u6790\u8207\u8a0e\u8ad6 \u56db\u3001\u5be6\u9a57\u7d50\u679c\u53ca\u8a0e\u8ad6 \u4ee5\u53ca\u7576\u524d\u55ae\u5143\u5c64\u7d1a\u7531\u524d\u81f3\u5f8c\u548c\u7531\u5f8c\u81f3\u524d\u6578\u4f86\u7684\u4f4d\u7f6e\u90fd\u7d0d\u5165\u8003\u616e\uff0c\u4e26\u5c07\u4ee5\u4e0a\u7684\u97fb\u5f8b\u968e\u5c64 \u5728\u5be6\u9a57\u6578\u64da\u4e0a\u986f\u793a\uff0c\u672c\u8ad6\u6587\u6240\u63d0\u51fa\u7684\u5c07\u53e5\u5b50\u6574\u53e5\u5408\u6210\u5728\u5c07\u5b57\u66ff\u63db\u56de\u539f\u672c\u7684\u5b57\uff0c\u78ba\u5be6 \u5176\u4e2d\u503c\u5f97\u6ce8\u610f\u7684\u662f\u97f3\u6a19\u4e2d\u7684[l]\u3001[m]\u3001[n]\u96d6\u7136\u5728\u97f3\u6a19\u5167\u662f\u540c\u500b\u6a23\u5b50\uff0c\u4f46\u5be6\u969b\u4e0a\u6839\u64da\u662f\u5426 \u5f9e\u5be6\u9a57\u5e73\u5747\u548c\u9577\u689d\u5716\u4f86\u770b\uff0c\u591a\u6578\u53e5\u5b50\u90fd\u986f\u793a\u6709\u5206\u6790\u6587\u8108\u5f8c\u7684\u7d50\u679c\u8f03\u597d\u3002\u4ee5\u4e0b\u5c07\u7528\u4e00\u500b\u6b63 \u53ef\u7531\u5206\u985e\u56de\u6b78\u6c7a\u7b56\u6a39(Classification and Regression Tree, CART) [13]\u5f97\u5230\u3002 (\u4e00)\u3001\u5be6\u9a57\u76ee\u7684\u8207\u8a9e\u6599\u5eab\u4ecb\u7d39 \u6bd4\u6bcf\u6bb5\u5b57\u5206\u958b\u5408\u6210\u5f8c\u5728\u5408\u4f75\u4f86\u7684\u597d\uff0c\u56e0\u70ba\u5176\u4fdd\u6709\u4e86\u6574\u53e5\u8a71\u7684\u6d41\u66a2\u6027\u800c\u975e\u50cf\u662f\u5404\u500b\u55ae\u5b57\u62fc \u5728\u6bcd\u97f3\u524d\uff0c\u770b\u4f3c\u76f8\u540c\u7684\u97f3\u6a19\u6703\u6709\u4e0d\u540c\u7684\u767c\u97f3\uff0c\u9019\u5728 ARPAbet \u6a19\u793a\u6cd5\u4e2d\u9700\u4ee5\u4e0d\u540c\u7684\u8a18\u865f \u5f0f\u7684\u4e2d\u82f1\u593e\u96dc\u53e5\u5b50\u4f5c\u70ba\u7bc4\u4f8b\u4f86\u89c0\u5bdf\u983b\u8b5c\u5716\u548c\u6587\u8108\u4e2d\u5982\u4f55\u77e5\u9053\u70ba\u4ec0\u9ebc\u4e0d\u5206\u6790\u524d\u5f8c\u6587\u6703\u5c0e \u85c9\u7531\u4ee5\u4e0a\u6587\u8108\u8cc7\u8a0a\u8a18\u9304\uff0c\u6211\u5011\u53ef\u7531 Label \u6a94\u4e2d\u770b\u51fa\u6574\u53e5\u8a71\u7684\u65b7\u8a5e\u662f\u5982\u4f55\u65b7\u7684\uff0c\u5728\u5716 \u63a5\u800c\u6210\u3002</td></tr></table>", |
| "html": null |
| } |
| } |
| } |
| } |
