| { | |
| "paper_id": "O14-1003", | |
| "header": { | |
| "generated_with": "S2ORC 1.0.0", | |
| "date_generated": "2023-01-19T08:04:35.419440Z" | |
| }, | |
| "title": "", | |
| "authors": [], | |
| "year": "", | |
| "venue": null, | |
| "identifiers": {}, | |
| "abstract": "", | |
| "pdf_parse": { | |
| "paper_id": "O14-1003", | |
| "_pdf_hash": "", | |
| "abstract": [], | |
| "body_text": [ | |
| { | |
| "text": "\u7a76\u7684\u5be6\u9a57\u7d50\u679c\uff0c\u76ee\u524d\u5e38\u7528\u7684\u516c\u958b\u60c5\u7dd2\u8a9e\u6599\u5eab\u5982\u5fb7\u8a9e\u7684 Berlin Database of Speech Emotion (EMO-DB) [3] \u3001FAU Aibo [4] \u548c Vera-Am-Mittag(VAM) [5] ", | |
| "cite_spans": [ | |
| { | |
| "start": 67, | |
| "end": 70, | |
| "text": "[3]", | |
| "ref_id": "BIBREF2" | |
| }, | |
| { | |
| "start": 81, | |
| "end": 84, | |
| "text": "[4]", | |
| "ref_id": "BIBREF3" | |
| }, | |
| { | |
| "start": 107, | |
| "end": 110, | |
| "text": "[5]", | |
| "ref_id": "BIBREF4" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u7814\u7a76\u8a9e\u97f3\u60c5\u7dd2\u8fa8\u8b58\u5c31\u5fc5\u9808\u5177\u6709\u60c5\u7dd2\u8a9e\u97f3\u8a9e\u6599\u5eab\u7528\u65bc\u5be6\u9a57\uff0c\u800c\u8a9e\u6599\u5eab\u7684\u6536\u9304\u6703\u76f4\u63a5\u5f71\u97ff\u7814", | |
| "sec_num": null | |
| } | |
| ], | |
| "back_matter": [], | |
| "bib_entries": { | |
| "BIBREF0": { | |
| "ref_id": "b0", | |
| "title": "Affective computing", | |
| "authors": [ | |
| { | |
| "first": "R", | |
| "middle": [ | |
| "W" | |
| ], | |
| "last": "Picard", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1997, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "R. W. Picard, Affective computing. MIT Press, 1997.", | |
| "links": null | |
| }, | |
| "BIBREF1": { | |
| "ref_id": "b1", | |
| "title": "Survey on speech emotion recognition: features, classification schemes, and databases", | |
| "authors": [ | |
| { | |
| "first": "M", | |
| "middle": [ | |
| "El" | |
| ], | |
| "last": "Ayadi", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [ | |
| "S" | |
| ], | |
| "last": "Kamel", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "F", | |
| "middle": [], | |
| "last": "Karray", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2011, | |
| "venue": "Pattern Recognition", | |
| "volume": "44", | |
| "issue": "", | |
| "pages": "572--587", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "M. El Ayadi, M. S. Kamel, and F. Karray, \"Survey on speech emotion recognition: features, classification schemes, and databases,\" in Pattern Recognition, vol. 44, pp. 572-587, Mar. 2011.", | |
| "links": null | |
| }, | |
| "BIBREF2": { | |
| "ref_id": "b2", | |
| "title": "A database of german emotional speech", | |
| "authors": [ | |
| { | |
| "first": "F", | |
| "middle": [], | |
| "last": "Burkhardt", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Paeschke", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Rolfes", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "W", | |
| "middle": [ | |
| "F" | |
| ], | |
| "last": "Sendlmeier", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "B", | |
| "middle": [], | |
| "last": "Weiss", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2005, | |
| "venue": "Proceedings of INTERSPEECH", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "1517--1520", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "F. Burkhardt, A. Paeschke, M. Rolfes, W. F. Sendlmeier, and B. Weiss, \"A database of german emotional speech,\" in Proceedings of INTERSPEECH, pp. 1517-1520, 2005.", | |
| "links": null | |
| }, | |
| "BIBREF3": { | |
| "ref_id": "b3", | |
| "title": "Automatic classification of emotion related user states in spontaneous children's speech", | |
| "authors": [ | |
| { | |
| "first": "S", | |
| "middle": [], | |
| "last": "Steidl", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2009, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "S. Steidl, Automatic classification of emotion related user states in spontaneous children's speech. PhD thesis, University of Erlangen-Nuremberg, 2009.", | |
| "links": null | |
| }, | |
| "BIBREF4": { | |
| "ref_id": "b4", | |
| "title": "The Vera am Mittag german audio-visual emotional speech database", | |
| "authors": [ | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Grimm", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "K", | |
| "middle": [], | |
| "last": "Kroschel", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "S", | |
| "middle": [], | |
| "last": "Narayanan", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2008, | |
| "venue": "Proceedings of IEEE International Conference on Multimedia and Expo", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "865--868", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "M. Grimm, K. Kroschel, and S. Narayanan, \"The Vera am Mittag german audio-visual emotional speech database,\" in Proceedings of IEEE International Conference on Multimedia and Expo, pp. 865-868, 2008.", | |
| "links": null | |
| }, | |
| "BIBREF5": { | |
| "ref_id": "b5", | |
| "title": "Crowdsourcing facial responses to online videos", | |
| "authors": [ | |
| { | |
| "first": "D", | |
| "middle": [], | |
| "last": "Mcduff", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "R", | |
| "middle": [], | |
| "last": "Kaliouby", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "R", | |
| "middle": [], | |
| "last": "Picard", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2012, | |
| "venue": "IEEE Transactions on Affective Computing", | |
| "volume": "3", | |
| "issue": "", | |
| "pages": "456--468", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "D. McDuff, R. Kaliouby, and R. Picard, \"Crowdsourcing facial responses to online videos,\" in IEEE Transactions on Affective Computing, vol. 3, pp. 456-468, 2012.", | |
| "links": null | |
| }, | |
| "BIBREF6": { | |
| "ref_id": "b6", | |
| "title": "The INTERSPEECH 2009 emotion challenge", | |
| "authors": [ | |
| { | |
| "first": "B", | |
| "middle": [], | |
| "last": "Schuller", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "S", | |
| "middle": [], | |
| "last": "Steidl", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Batliner", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2009, | |
| "venue": "Proceedings of INTERSPEECH", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "312--315", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "B. Schuller, S. Steidl, and A. Batliner, \"The INTERSPEECH 2009 emotion challenge,\" in Proceedings of INTERSPEECH, pp. 312-315, 2009.", | |
| "links": null | |
| }, | |
| "BIBREF7": { | |
| "ref_id": "b7", | |
| "title": "The effect of class distribution on classifier learning: An empirical study", | |
| "authors": [ | |
| { | |
| "first": "G", | |
| "middle": [], | |
| "last": "Weiss", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "F", | |
| "middle": [], | |
| "last": "Provost", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2001, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "G. Weiss and F. Provost, \"The effect of class distribution on classifier learning: An empirical study,\" tech. rep., Department of Computer Science, Rutgers University, 2001.", | |
| "links": null | |
| }, | |
| "BIBREF8": { | |
| "ref_id": "b8", | |
| "title": "Acoustic emotion recognition: A benchmark comparison of performances", | |
| "authors": [ | |
| { | |
| "first": "B", | |
| "middle": [], | |
| "last": "Schuller", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "B", | |
| "middle": [], | |
| "last": "Vlasenko", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "F", | |
| "middle": [], | |
| "last": "Eyben", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "G", | |
| "middle": [], | |
| "last": "Rigoll", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Wendemuth", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2009, | |
| "venue": "Proceedings of IEEE Workshop on of Automatic Speech Recognition Understanding (ASRU)", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "552--557", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "B. Schuller, B. Vlasenko, F. Eyben, G. Rigoll, and A. Wendemuth, \"Acoustic emotion recognition: A benchmark comparison of performances,\" in Proceedings of IEEE Workshop on of Automatic Speech Recognition Understanding (ASRU), pp. 552-557, 2009.", | |
| "links": null | |
| }, | |
| "BIBREF9": { | |
| "ref_id": "b9", | |
| "title": "OpenEAR introducing the munich open-source emotion and affect recognition toolkit", | |
| "authors": [ | |
| { | |
| "first": "F", | |
| "middle": [], | |
| "last": "Eyben", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Wollmer", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "B", | |
| "middle": [], | |
| "last": "Schuller", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2009, | |
| "venue": "Proceedings of ACII International Conference on Affective Computing and Intelligent Interaction and Workshops", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "1--6", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "F. Eyben, M. Wollmer, and B. Schuller, \"OpenEAR introducing the munich open-source emotion and affect recognition toolkit,\" in Proceedings of ACII International Conference on Affective Computing and Intelligent Interaction and Workshops, 2009, pp. 1-6.", | |
| "links": null | |
| }, | |
| "BIBREF10": { | |
| "ref_id": "b10", | |
| "title": "Weka: Practical machine learning tools and techniques with java implementations", | |
| "authors": [ | |
| { | |
| "first": "I", | |
| "middle": [ | |
| "H" | |
| ], | |
| "last": "Witten", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "E", | |
| "middle": [], | |
| "last": "Frank", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "L", | |
| "middle": [], | |
| "last": "Trigg", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Hall", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "G", | |
| "middle": [], | |
| "last": "Holmes", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "S", | |
| "middle": [ | |
| "J" | |
| ], | |
| "last": "Cunningham", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1999, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "I. H. Witten, E. Frank, L. Trigg, M. Hall, G. Holmes, and S. J. Cunningham, \"Weka: Practical machine learning tools and techniques with java implementations,\" 1999.", | |
| "links": null | |
| } | |
| }, | |
| "ref_entries": { | |
| "TABREF0": { | |
| "num": null, | |
| "type_str": "table", | |
| "html": null, | |
| "text": "\u751f\u6c23 127 \u53e5\u3001\u7121\u804a 81 \u53e5\u3001\u5641\u5fc3 46 \u53e5\u3001\u5bb3\u6015 69 \u53e5\u3001\u958b\u5fc3 71 \u53e5\u3001\u50b7\u5fc3 62 \u53e5\u548c\u4e2d\u6027 79 \u53e5 \u78bc\uff0c\u5206\u5225\u70ba\u7b2c\u4e00\u78bc\u8a9e\u8a00 ID\u3001\u7b2c\u4e8c\u4e09\u78bc\u8a9e\u8005 ID\uff0c\u7b2c\u56db\u78bc\u60c5\u7dd2 ID\uff0c\u7b2c\u4e94\u516d\u4e03\u78bc\u6587 \u672c ID\uff0c\u5982 H03At01 \u97f3\u6a94\u5247\u70ba\u5ba2\u5bb6\u8a9e\u3001\u8a9e\u8005 03\u3001\u60c5\u7dd2\u70ba\u751f\u6c23\u548c t01 \u6587\u672c\u3002\u63a5\u8457\u5206\u5225\u5c0d\u4e09\u7a2e \u8a9e\u8a00\u5404\u627e 5 \u4f4d\u7537\u6027\u4ee5\u53ca 5 \u4f4d\u5973\u6027\u4f5c\u70ba\u9304\u97f3\u8005\u3002\u76f8\u540c\u7684\u6211\u5011\u5728\u9304\u88fd\u5b8c\u5f8c\u4e5f\u8acb\u4e86 10 \u4f4d\u6e2c\u8a66\u8005", | |
| "content": "<table><tr><td>\u3002[6]\u63d0\u53ca\u5728\u6536\u96c6\u60c5\u611f\u8cc7\u6599\u6642\uff0c\u4eba \u5011\u8655\u65bc\u5be6\u9a57\u5ba4\u7b49\u975e\u5e73\u5e38\u7fd2\u6163\u7684\u5834\u5408\u6240\u6536\u96c6\u5230\u7684\u60c5\u611f\u8cc7\u6599\u5be6\u969b\u4e0a\u6703\u8207\u771f\u5be6\u4e16\u754c\u8868\u9054\u51fa\u7684\u6709 \u6240\u5dee\u7570\uff0c\u56e0\u6b64\u60c5\u7dd2\u8a9e\u97f3\u8a9e\u6599\u5eab\u7684\u9304\u88fd\u5927\u81f4\u80fd\u5206\u70ba\u5169\u985e\uff1b\u7b2c\u4e00\u985e\u70ba\u5f15\u5c0e\u6027\u9304\u97f3\uff0c\u6b64\u985e\u591a\u5728 \u5be6\u9a57\u5ba4\u6216\u8005\u662f\u9304\u97f3\u5ba4\u4e2d\u9304\u88fd\uff0c\u5982\u5fb7\u8a9e\u7684 EMO-DB\uff0c\u5176\u900f\u904e\u5728\u9304\u97f3\u5ba4\u9304\u88fd\u4e26\u4e14\u6709\u4e09\u4f4d\u8a9e\u8a00 \u5b78\u5c08\u5bb6\u5728\u65c1\u76e3\u7763\u8207\u6307\u5c0e\uff0c\u5982\u6b64\u53ef\u4ee5\u78ba\u4fdd\u9304\u88fd\u51fa\u4e7e\u6de8\u4e26\u4e14\u60c5\u7dd2\u8868\u9054\u5ea6\u548c\u5dee\u7570\u6027\u9ad8\u7684\u60c5\u7dd2\u8a9e \u6599\u3002\u6b64\u7a2e\u9304\u88fd\u65b9\u5f0f\u8f03\u9700\u8981\u8a9e\u8005\u7684\u8868\u9054\u80fd\u529b\uff0c\u5728\u8a9e\u8005\u8868\u9054\u80fd\u529b\u8f03\u5dee\u7684\u60c5\u6cc1\u4e0b\u7121\u6cd5\u5728\u9304\u97f3\u5ba4 \u9304\u88fd\u51fa\u542b\u6709\u8db3\u5920\u60c5\u7dd2\u8868\u9054\u7684\u8a9e\u53e5;\u76f8\u5c0d\u7684\u7b2c\u4e8c\u7a2e\u5247\u662f\u81ea\u767c\u6027\u7684\u60c5\u7dd2\u8868\u9054\u8a9e\u53e5\uff0c\u5982 FAU Aibo \u548c VAM\uff0cFAU Aibo \u900f\u904e\u9304\u88fd\u5c0f\u5b69\u5b50\u8207\u6a5f\u5668\u4eba\u7684\u4e92\u52d5\uff0c\u4fbf\u53ef\u4ee5\u76f4\u63a5\u6536\u96c6\u5230\u5c0f\u5b69\u5b50\u81ea \u7136\u8868\u9054\u7684\u60c5\u7dd2\u8a9e\u53e5\u3002\u800c VAM \u5247\u662f\u5f9e\u8ac7\u8a71\u6027\u7bc0\u76ee\u4e2d\u7bc0\u9304\u7247\u6bb5\u518d\u5f9e\u4e2d\u6311\u9078\u51fa\u60c5\u611f\u8f03\u70ba\u6d3b\u8e8d \u7684\u53e5\u5b50\uff0c\u5982\u6b64\u6536\u9304\u7684\u8a9e\u6599\u5eab\u4e5f\u8f03\u81ea\u7136\u5448\u73fe\u3002 \u5728\u8a9e\u97f3\u60c5\u7dd2\u8fa8\u8b58\u4e2d\u63d0\u51fa\u4f5c\u70ba\u57fa\u6e96\u5be6\u9a57\u7684\u6709\u5169\u8005\u3002\u7531\u65bc\u60c5\u611f\u8fa8\u8b58\u76f8\u5c0d\u65bc\u8a9e\u97f3\u8fa8\u8b58\u53ca\u8a9e \u8005 \u8fa8 \u8b58 \u70ba \u8f03 \u65b0 \u7684 \u9818 \u57df \uff0c \u56e0 \u6b64 \u8f03 \u65e9 \u4e26 \u6c92 \u6709 \u4e00 \u500b \u516c \u8a8d \u7684 \u57fa \u6e96 \u53ef \u4f9b \u53c3 \u8003 \u6bd4 \u8f03 \uff0c \u6240 \u4ee5 INTERSPEECH 2009 Emotion Challenge[7]\u4fbf\u63a1\u7528\u4e86\u5177\u6709\u660e\u78ba\u8a13\u7df4\u96c6\u53ca\u6e2c\u8a66\u96c6\u7684 FAU Aibo \u60c5\u7dd2\u8a9e\u6599\u5eab\uff0c\u5728\u524d\u7aef\u4f7f\u7528\u904e\u96f6\u7387(zero crossing rate, ZCR)\u3001\u80fd\u91cf\u3001\u57fa\u983b\u3001\u6cdb\u97f3\u566a\u97f3\u6bd4 (harmonics-to-noise, HNR)\u548c\u6885\u723e\u5012\u983b\u8b5c\u7cfb\u6578\u3002\u70ba\u4e86\u89e3\u6c7a FAU Aibo \u4e94\u985e\u8cc7\u6599\u91cf\u4e2d\u6709\u5169\u985e \u4f54\u4e86\u8fd1 80%\u7684\u8cc7\u6599\u6578\u91cf\u56b4\u91cd\u4e0d\u5e73\u5747\u7684\u72c0\u6cc1\uff0c\u4f7f\u7528 SMOTE(Synthetic Minority Oversampling Technology)[8]\u65b9\u6cd5\u4f86\u589e\u52a0\u6578\u91cf\u8f03\u5c11\u7684\u985e\u5225\u7684\u6a23\u672c\u6578\uff0c\u5f8c\u7aef\u5247\u4ee5\u96b1\u85cf\u5f0f\u99ac\u53ef\u592b\u6a21\u578b\u4ee5\u53ca\u652f \u6301\u5411\u91cf\u6a5f\u505a\u70ba\u8fa8\u8b58\u5668\uff0c\u4ee5\u6b64\u505a\u70ba FAU Aibo \u7684\u4e00\u500b\u5be6\u9a57\u57fa\u6e96\u3002[9]\u5be6\u9a57 9 \u500b\u60c5\u7dd2\u8a9e\u6599\u5eab\u5206 \u5225\u70ba Airplane Behaviour Corpus (ABC)\u3001Audiovisual Interest Corpus (AVIC) \u3001Danish Emotional Speech (DES)\u3001EMO-DB\u3001eNTERFACE\u3001Sensitive Artificial Listener (SAL)\u3001 SmartKom\u3001Speech Under Simulated and Actual Stress (SUSAS)\u3001\u548c VAM \u7b49\u8cc7\u6599\u5eab\uff0c\u9019\u4e9b \u8a9e\u6599\u5eab\u5206\u5225\u6709\u4e0d\u540c\u7684\u9304\u97f3\u689d\u4ef6\uff0c\u5305\u542b\u5f15\u5c0e\u5f0f\u9304\u97f3\u3001\u81ea\u767c\u6027\u8a9e\u97f3\u8207\u5404\u7a2e\u7684\u9304\u97f3\u74b0\u5883\u7b49\u7b49\u3002 \u5be6\u9a57\u5206\u70ba\u5169\u7a2e\uff0c\u7b2c\u4e00\u7a2e\u70ba\u64f7\u53d6\u4e00\u822c\u97f3\u6846\u7279\u5fb5\u4e26\u642d\u914d\u96b1\u85cf\u5f0f\u99ac\u53ef\u592b\u6a21\u578b/\u9ad8\u65af\u6df7\u5408\u6a21\u578b\u505a\u70ba \u5f8c\u7aef\u8fa8\u8b58\u5668\uff0c\u53e6\u4e00\u7a2e\u70ba\u622a\u53d6\u5927\u91cf\u7684\u8072\u5b78\u8cc7\u8a0a\u4e26\u7d93\u7531\u6cdb\u51fd\u5c07\u97f3\u6846\u7279\u5fb5\u8f49\u63db\u70ba\u53e5\u5b50\u7279\u5fb5\uff0c\u5982 \u6b64\u4ee5\u65b9\u4fbf\u5c0d\u61c9\u81f3\u5f8c\u7aef\u652f\u6301\u5411\u91cf\u6a5f\u7684\u8a13\u7df4\u8207\u8fa8\u8b58\uff0c\u6b64\u7bc7\u53ef\u8996\u70ba\u5404\u60c5\u7dd2\u8a9e\u6599\u5eab\u63a1\u7528\u8072\u5b78\u6a21\u578b \u6642\u7684\u57fa\u6e96\u5be6\u9a57\u53c3\u8003\u3002 \u76ee\u524d\u5e38\u898b\u7684\u516c\u958b\u8a9e\u97f3\u60c5\u7dd2\u8fa8\u8b58\u7684\u8a9e\u6599\u5eab\u4e26\u6c92\u6709\u4ee5\u53f0\u7063\u672c\u571f\u8a9e\u8a00\u9304\u7f6e\u800c\u6210\uff0c\u800c\u8a9e\u97f3\u4e2d \u5305\u542b\u4e86\u8a9e\u8a00\u3001\u8a9e\u8005\u3001\u6587\u672c\u3001\u8154\u8abf\u7b49\u7b49\u7684\u8cc7\u8a0a\uff0c\u9019\u4e9b\u7686\u6703\u5f71\u97ff\u8a9e\u97f3\u60c5\u7dd2\u8fa8\u8b58\u7684\u6e96\u78ba\u5ea6\u3002\u56e0 \u6b64\u70ba\u4e86\u63d0\u4f9b\u4e00\u500b\u8f03\u9069\u5408\u7528\u65bc\u53f0\u7063\u8a9e\u8a00\u7684\u60c5\u7dd2\u8a9e\u6599\u5eab\uff0c\u6211\u5011\u81ea\u884c\u9304\u88fd\u4e86\u53f0\u7063\u6700\u666e\u904d\u7684\u570b\u8a9e\u3001 \u53f0\u8a9e\u53ca\u5ba2\u5bb6\u8a9e\u4e09\u7a2e\u8a9e\u8a00\u7684\u60c5\u7dd2\u8a9e\u6599\u5eab\u3002\u6211\u5011\u4eff\u9020 EMO-DB \u9304\u88fd\u65b9\u5f0f\u9304\u88fd\uff0c\u5176\u4e2d\u76f8\u540c\u7684\u4e03 \u7a2e\u60c5\u7dd2\u3001\u4e94\u4f4d\u7537\u6027\u4e94\u4f4d\u5973\u6027\u5171\u5341\u4f4d\u8a9e\u8005\u4ee5\u53ca\u5341\u53e5\u6587\u672c\uff0c\u9304\u88fd\u5b8c\u5f8c\u9032\u884c\u4eba\u5de5\u8fa8\u8b58\u6e2c\u9a57\u3002\u6700 \u5f8c\u518d\u63a1\u7528\u8207[9]\u76f8\u540c\u7684\u57fa\u6e96\u5be6\u9a57\u8a2d\u5b9a\uff0c\u4f5c\u70ba\u6211\u5011\u8a9e\u6599\u5eab\u7684\u57fa\u6e96\u8fa8\u8b58\u7387\u3002 \u4ee5\u4e0b\u70ba\u6574\u7bc7\u8ad6\u6587\u7684\u67b6\u69cb\uff0c\u7b2c\u4e00\u7bc0\u70ba\u7dd2\u8ad6\uff0c\u7b2c\u4e8c\u7bc0\u8a73\u7d30\u4ecb\u7d39\u6211\u5011\u4eff\u9020\u7684\u5fb7\u8a9e\u8a9e\u6599\u5eab EMO-DB\uff0c\u7b2c\u4e09\u7bc0\u8a73\u5217\u6211\u5011\u81ea\u884c\u9304\u88fd\u7684\u53f0\u7063\u8a9e\u6599\u5eab\u8cc7\u8a0a\uff0c\u7b2c\u56db\u7bc0\u70ba\u53f0\u7063\u8a9e\u6599\u5eab\u7684\u57fa\u6e96\u5be6 \u4e8c\u3001\u73fe\u6709\u516c\u958b\u8a9e\u6599\u5eab \u6211\u5011\u9304\u88fd\u7684\u8a9e\u6599\u5eab\u53c3\u8003\u5df2\u5ee3\u6cdb\u5be6\u9a57\u65bc\u8a9e\u97f3\u60c5\u7dd2\u7814\u7a76\u7684\u5fb7\u8a9e\u8a9e\u6599\u5eab Berlin Database of Emotion Speech(EMO-DB)\u3002EMO-DB \u7531 10 \u4f4d\u5fb7\u8a9e\u6bcd\u8a9e\u8a9e\u8005\u6240\u9304\u88fd\u800c\u6210\uff0c\u7d93\u7531\u9ea5\u514b\u98a8\u9304 \u88fd 48kHz \u5f8c\u964d\u70ba 16kHz\u3002\u9304\u97f3\u74b0\u5883\u5728\u9694\u97f3\u826f\u597d\u53ca\u5177\u6709\u9ad8\u54c1\u8cea\u9304\u97f3\u8a2d\u5099\u7684\u9304\u97f3\u5ba4\uff0c\u56e0\u6b64\u8868 \u9054\u7684\u60c5\u7dd2\u4e26\u975e\u81ea\u7136\u5448\u73fe\u3002\u8a9e\u8005\u7531 3 \u4f4d\u8a9e\u8a00\u5b78\u5c08\u5bb6\u5f9e 40 \u4f4d\u65b0\u805e\u64ad\u5831\u54e1\u4e2d\u6311\u9078\uff0c\u5728\u9762\u8a66\u6642\u6bcf \u4eba\u5c0d\u6bcf\u7a2e\u60c5\u7dd2\u900f\u904e\u9ea5\u514b\u98a8\u9304\u88fd\u4e00\u53e5\u6e2c\u8a66\u53e5\uff0c\u518d\u5c07\u9019\u4e9b\u53e5\u5b50\u4ea4\u4e88\u4e09\u4f4d\u8a9e\u8a00\u5c08\u5bb6\uff0c\u5c08\u5bb6\u6839\u64da \u81ea\u7136\u5ea6\u53ca\u53ef\u8fa8\u8b58\u5ea6\u6311\u9078\u51fa\u7537\u5973\u5404 5 \u4f4d\u4f5c\u70ba\u9304\u97f3\u8a9e\u8005\u3002\u8a9e\u6599\u5eab\u5171\u63a1\u7528 7 \u7a2e\u60c5\u7dd2\u5206\u5225\u70ba\u4e2d\u6027\u3001 \u751f\u6c23\u3001\u5bb3\u6015\u3001\u958b\u5fc3\u3001\u50b7\u5fc3\u3001\u5641\u5fc3\u548c\u7121\u804a\uff0c\u6587\u672c\u8a2d\u8a08\u70ba\u4e94\u53e5\u9577\u53e5\u53ca\u4e94\u53e5\u77ed\u53e5\uff0c\u6587\u672c\u5167\u5bb9\u70ba \u65e5\u5e38\u751f\u6d3b\u4e2d\u7684\u53e5\u5b50\u4e26\u4e14\u4e0d\u5e36\u6709\u4efb\u4f55\u60c5\u7dd2\u8a5e\u5f59\uff0c\u5982\u6b64\u4f7f\u8a9e\u8005\u5728\u9304\u97f3\u6642\u80fd\u8f03\u81ea\u7136\u7684\u8868\u9054\u8a9e\u53e5\uff0c \u4e14\u5728\u60c5\u7dd2\u8868\u9054\u904e\u7a0b\u4e2d\u4e0d\u6703\u53d7\u5230\u6587\u672c\u7684\u5167\u5bb9\u800c\u5f71\u97ff\u3002\u9304\u88fd\u6642\u8a9e\u8005\u5c0d\u76f8\u540c\u4e00\u53e5\u5206\u5225\u4ee5\u4e03\u7a2e\u60c5 \u7dd2\u8868\u9054\uff0c\u56e0\u6b64\u6bcf\u4f4d\u8a9e\u8005\u81f3\u5c11\u6536\u9304 70 \u53e5\uff0c\u6574\u500b\u8a9e\u6599\u5eab\u81f3\u5c11\u5305\u542b 700 \u53e5\uff0c\u6700\u5f8c\u5168\u90e8\u5171\u9304\u88fd\u7d04 800 \u53e5\uff0c\u5176\u4e2d\u6709\u90e8\u5206\u53e5\u5b50\u4fdd\u7559\u4e8c\u81f3\u4e09\u500b\u7248\u672c\u3002\u70ba\u4e86\u78ba\u4fdd\u8a9e\u53e5\u7684\u54c1\u8cea\uff0c\u8acb 20 \u4f4d\u6e2c\u8a66\u8005\u4ee5\u96a8 \u6a5f\u8f2a\u64ad\u7684\u65b9\u5f0f\u5c07\u6bcf\u4e00\u53e5\u505a\u4e03\u9078\u4e00\u7684\u4eba\u5de5\u8fa8\u8b58\uff0c\u4e26\u4e14\u5c0d\u53e5\u5b50\u7684\u81ea\u7136\u5ea6\u4ee5 1 \u81f3 100 \u8a55\u5206\uff0c\u6e2c \u9a57\u7d50\u675f\u5f8c\u6368\u68c4\u8fa8\u8b58\u7387\u4f4e\u65bc 80%\u548c\u81ea\u7136\u5ea6\u4f4e\u65bc 60%\u7684\u53e5\u5b50\uff0c\u7be9\u9078\u5f8c\u4e03\u7a2e\u60c5\u7dd2\u7684\u53e5\u6578\u5206\u5225\u70ba \u5171 535 \u53e5\u3002 \u4e09\u3001\u53f0\u7063\u60c5\u7dd2\u8a9e\u6599\u5eab \u70ba\u4e86\u7814\u7a76\u8de8\u8a9e\u8a00\u7684\u60c5\u7dd2\u8a9e\u97f3\u8fa8\u8b58\u5be6\u9a57\uff0c\u6211\u5011\u4eff\u7167\u4e86 EMO-DB \u7684\u8a2d\u8a08\u65b9\u5f0f\u9304\u88fd\u4e86\u81ea\u5df1 \u7684\u60c5\u7dd2\u8a9e\u6599\u5eab\uff0c\u5206\u5225\u7528\u76f8\u540c\u7684\u4e03\u7a2e\u60c5\u7dd2\u8868\u9054\u4e94\u53e5\u9577\u53e5\u4ee5\u53ca\u4e94\u53e5\u77ed\u53e5\uff0c\u6bcf\u7a2e\u8a9e\u8a00\u5404\u9304\u88fd 700 \u53e5\uff0c\u8868\u4e00\u5217\u51fa\u6211\u5011\u7684\u5341\u53e5\u6587\u672c\uff0c\u8868\u4e8c\u8207\u8868\u4e09\u5206\u5225\u70ba\u53f0\u8a9e\u8207\u5ba2\u5bb6\u8a9e\u7684\u767c\u97f3\u6587\u672c\u3002\u6211\u5011\u9078\u64c7\u53f0 \u7063\u5e38\u898b\u7684\u4e09\u7a2e\u8a9e\u8a00-\u570b\u8a9e\u3001\u53f0\u8a9e\u53ca\u5ba2\u5bb6\u8a9e-\u4f5c\u70ba\u6211\u5011\u7684\u53f0\u7063\u9304\u88fd\u8a9e\u8a00\uff0c\u7531\u65bc\u5ba2\u5bb6\u8a9e\u6709\u4e0d\u540c\u8154 \u8abf\uff0c\u56e0\u6b64\u6211\u5011\u6311\u9078\u8a9e\u8005\u6642\u9078\u64c7\u6700\u70ba\u666e\u904d\u7684\u56db\u7e23\u8154\u8207\u6d77\u9678\u8154\u4f5c\u70ba\u6211\u5011\u7684\u9304\u88fd\u8154\u8abf\u3002\u9304\u97f3\u74b0\u5883 \u70ba\u4e2d\u5c71\u5927\u5b78\u96fb\u8cc7\u5927\u6a13 F5017B \u5be6\u9a57\u5ba4\uff0c\u9304\u97f3\u9ea5\u514b\u98a8\u4f7f\u7528 ATH-AT9942\uff0c\u9304\u97f3\u4ecb\u9762\u5361\u70ba Sound Blaster X-Fi Surround 5.1\uff0c\u53d6\u6a23\u7387\u70ba 16KHz\uff0c\u9304\u97f3\u6642\u8a9e\u8005\u8207\u9ea5\u514b\u98a8\u9593\u8ddd\u7d04 20 \u516c\u5206\u3002\u6a94\u6848\u547d \u540d\u683c\u5f0f\u70ba 7 \u505a\u4eba\u5de5\u8fa8\u8b58\uff0c\u4f46\u6211\u5011\u7684\u6e2c\u9a57\u4e2d\u4e26\u6c92\u6709\u81ea\u7136\u5ea6\u7684\u8a55\u5206\u3002\u6211\u5011\u5f9e\u4e2d\u4fdd\u7559\u8fa8\u8b58\u7387 60%\u4ee5\u4e0a\u7684\u53e5\u5b50\uff0c \u8868\u56db\u5217\u51fa\u4e09\u7a2e\u8a9e\u8a00\u6311\u9078\u5f8c\u5404\u60c5\u7dd2\u6240\u5269\u9918\u7684\u53e5\u6578\uff0c\u5716\u4e00\u8868\u793a\u4e09\u7a2e\u60c5\u7dd2\u5404\u500b\u4eba\u5de5\u8fa8\u8b58\u7387\u7684\u53e5\u6578\uff0c \u7531\u5716\u4e2d\u53ef\u770b\u51fa\u591a\u6578\u9304\u88fd\u7684\u4eba\u5de5\u6e96\u78ba\u5ea6\u5728 60%\u4ee5\u4e0a\uff0c\u6240\u4ee5\u6211\u5011\u7684\u60c5\u7dd2\u8a9e\u6599\u5eab\u5177\u6709\u4e00\u5b9a\u7684\u60c5\u7dd2 \u8fa8\u5225\u5ea6\u3002\u6240\u6709\u7684\u9304\u97f3\u8005\u548c\u6e2c\u8a66\u8005\u7686\u70ba\u5927\u5b78\u751f\u6216\u662f\u7814\u7a76\u751f\uff0c\u4e26\u4e14\u975e\u8a9e\u8a00\u5b78\u6216\u8005\u97f3\u6a02\u5b78\u7cfb\u5b78 \u7684\u57fa\u6e96\u7279\u5fb5\u96c6\u3002 \u751f\u3002 \u8868\u4e00\u3001\u53f0\u7063\u60c5\u7dd2\u8a9e\u6599\u5eab 10 \u53e5\u6587\u672c \u6587\u672c ID \u6587\u672c t01 \u4f60\u7684\u65e9\u9910\u653e\u5728\u684c\u4e0a t02 \u665a\u4e0a\u4ed6\u6709\u4e00\u500b\u7d04\u6703 t03 \u6700\u8fd1\u5e38\u5e38\u7761\u4e0d\u98fd t04 \u7b49\u4e0b\u4e00\u8d77\u53bb\u9910\u5ef3\u5403\u98ef t05 \u904e\u5e74\u8981\u8cb7\u4e00\u4ef6\u65b0\u8863\u670d t06 \u9019\u79ae\u62dc\u653e\u5047\uff0c\u8981\u8ddf\u670b\u53cb\u4e00\u8d77\u51fa\u53bb\u73a9 t07 \u8868\u4e09\u3001\u5ba2\u5bb6\u8a9e\u5c0d\u61c9\u6587\u672c \u6587\u672c ID \u6587\u672c t01 (\u310b\u3127\u311a\u02c7)(\u310d\u311f)\u5492\u4e5d(\u3105\u3129\u3125)\u515c\u63cd\u8f5f t02 \u52d5\u614b\u7279\u5fb5\u8a08\u7b97\u70ba delta regression\uff0c\u4fc2\u6578\u5dee\u516c\u5f0f\u63a1\u7528 Hidden Markov Model Tool Kit \u5b9a\u7fa9\u5982\u7b97 \u8868\u516b\u3001\u53f0\u8a9e\u57fa\u6e96\u5be6\u9a57\u6df7\u6dc6\u77e9\u9663 \u8868\u4e94\u300156 \u500b\u4f4e\u968e\u53c3\u6578 Feature Group Feature in Group # of LLD Raw signal Zero-crossing-rate 1 \u5f0f(3)\uff0c\u5176\u4e2d W \u8a2d\u70ba 2\u3002 \u7b54\u6848 \u751f\u6c23 \u7121\u804a \u5641\u5fc3 \u5bb3\u6015 \u958b\u5fc3 \u4e2d\u6027 \u50b7\u5fc3 \u6e96\u78ba\u7387 \u7d50\u679c (3) \u751f\u6c23 45 1 8 0 14 0 5 61.6 (\u3124)\u88dc\u58d3(\u310d\u4e00\u02c7)\u6b66\u4f9d\u70b8\u53f3(\u3108\u4e00) t03 \u8ffd(\u310e\u3127\u3125)\u5bf5\u5bf5\u96d6\u67d0\u5831 t04 Signal energy Logarithm 1 Pitch \u7121\u804a 0 43 0 0 0 7 5 78.2 Fundamental frequency F0 in Hz via Cepstrum and Autocorrelation (ACF). Exponentially smoothed F0 2 \u5641\u5fc3 6 4 27 2 4 8 4 49.1 (den)\u54c8(\u310e\u3127\u3125)\u54c8(\u310f\u4e00)\u8822\u5802(\u3115\u02d9)\u7ffb t05 envelope. 4.3 \u5be6\u9a57 \u5bb3\u6015 7 1 4 39 6 2 2 63.9 \u52fe\u8f3e(\u311b\u4e00)\u57cb\u5104\u826f\u4f08\u8cde\u592b t06 (\u310c\u4e00\u311a\u02cb)\u79ae\u63b0(\u3105\u4e00\u3125)(\u310c\u3127\u3120)\uff0c(\u311b\u4e00)\u8001\u54c1 (\u4e00\u3129\u02ca)(\u310e\u3127\u3125)\u54c8\u7c97(\u310f\u4e00)(\u310c\u3127\u3120) Voice Quality Probability of voicing 1 \u958b\u5fc3 12 5 4 3 24 1 9 41.4 \u6839\u64da\u8868\u4e94\u8207\u8868\u516d\u7684 6552 \u7dad\u5927\u578b\u7279\u5fb5\u96c6\u642d\u914d\u652f\u6301\u5411\u91cf\u6a5f\u4f5c\u70ba\u570b\u8a9e\u3001\u53f0\u8a9e\u548c\u5ba2\u5bb6\u8a9e\u4e09\u7a2e Energy in bands 0-250 Hz, 0-650 Hz, 250-650 Hz, 1-4 kHz 25%, 50%, 75%, 90% roll-off point, centroid, flux, \u8a9e\u8a00\u7684\u57fa\u6e96\u5be6\u9a57\uff0c\u5c0d\u6bcf\u500b\u8a9e\u8a00\u7684 10 \u4f4d\u8a9e\u8005\u9032\u884c LOSO \u5be6\u9a57\u5f8c\u5206\u5225\u5f97\u5230\u570b\u8a9e 68.5%\u3001\u53f0\u8a9e \u50b7\u5fc3 0 16 8 3 1 25 3 44.6 \u6628\u5929\u65e9\u4e0a\u51fa\u9580\u7684\u6642\u5019\uff0c\u5916\u5957\u88ab\u9264\u5b50\u52fe\u5230 t08 \u4eca\u5929\u4e00\u6574\u5929\u90fd\u6c92\u5403\u6771\u897f\u597d\u9913\u5594 t09 \u9019\u73ed\u706b\u8eca\u4eba\u5f88\u591a\uff0c\u5f88\u591a\u4eba\u90fd\u6c92\u4f4d\u5b50\u5750 t10 \u65e9\u4e0a\u53bb\u9a0e\u8173\u8e0f\u8eca\uff0c\u4e0b\u5348\u6563\u6b65\u53bb\u8cb7\u6771\u897f \u8868\u4e8c\u3001\u53f0\u8a9e\u5c0d\u61c9\u6587\u672c \u6587\u672c ID \u6587\u672c t01 \u4f60\u7684\u65e9\u9813\u653e\u5728\u684c\u9802 t02 \u6697\u6642\u4f9d\u6709\u4e00\u500b\u7d04\u6703 t03 \u6700\u8fd1\u5b9a\u5b9a\u774f\u4e0d\u98fd t04 \u7b49\u54a7\u505a\u4f19\u53bb\u9910\u5ef3\u98df\u98ef t05 \u919c\u6bd4\u5247\u640d(\u311b\u4e00)\u8655(\u3107\u3128\u3123\u02c7)(\u311f)\u6b7b(\u310f\u3127\u3121)\uff0c\u6b6a\u638f(\u3105 Spectral and rel. pos. of spectrum max. and min. & 12 12 50.7%\u548c\u5ba2\u5bb6\u8a9e 58.5%\u8fa8\u8b58\u7387\u3002\u8868\u4e03\u81f3\u8868\u4e5d\u5206\u5225\u5217\u51fa\u570b\u8a9e\u3001\u53f0\u8a9e\u548c\u5ba2\u5bb6\u8a9e\u4e09\u7a2e\u8a9e\u8a00\u7684\u8a73\u7d30 \u4e2d\u6027 16 13 6 6 10 2 11 17.2 t07 \u3128\u3123\u02c7)(\u310d\u3127\u3121\u02ca)(\u311f\u02c7)(\u310d\u3127\u3121\u02ca)\u8c46 t08 (\u310d\u3127\u3123\u02c7)(\u3105\u3128\u4e00\u02ca)\u9006(\u310d\u3128\u4e00\u02ca)\u9006(\u311f\u02c7)\u6731\u67d0\u5962\u61c2\u5e2d\u558a \u7aaf\u5594 t09 (\u310c\u3127\u311a\u02cb)(\u3105\u3127\u3125\u02ca)(\u3108\u3121\u02cb)\u67e5\u64f0\u61c2(\u3109\u3121\u02ca)\uff0c \u61c2(\u3109\u3121\u02ca)\u64f0\u6731\u67d0(\u3128\u4e00)(\u311f\u02cb)\u5f8c\u4ec7 t10 \u5247\u640d(\u310f\u4e00)(\u310e\u4e00\u02c7)\u5e2b\u6717\u67e5\uff0c(\u310f\u311a\u02c7)\u79df\u4e09(\u3105\u3128) (\u310f\u4e00)\u8cb7\u61c2\u5e2d \u8868\u56db\u3001\u53f0\u7063\u60c5\u7dd2\u8a9e\u6599\u5eab\u5404\u8a9e\u8a00\u7be9\u9078\u5f8c\u53e5\u6578 \u60c5\u7dd2 \u570b\u8a9e \u53f0\u8a9e \u5ba2\u5bb6\u8a9e \u8de8\u8a9e\u8a00\u7e3d\u6578 \u751f\u6c23 58 73 62 193 \u7121\u804a 66 55 59 180 \u5716\u4e00\u3001\u5404\u4eba\u5de5\u8fa8\u8b58\u7387\u7684\u53e5\u6578 \u56db\u3001\u57fa\u6e96\u5be6\u9a57 Mel-spectrum & Band 1-26 & 26 Cepstral & MFCC 0-12 & 13 Mel-spectrum Band 1-26 26 Cepstral MFCC 0-12 \u8fa8\u8b58\u7d50\u679c\u3002 \u8868\u4e5d\u3001\u5ba2\u5bb6\u8a9e\u57fa\u6e96\u5be6\u9a57\u6df7\u6dc6\u77e9\u9663. \u5f9e\u8868\u4e03\u81f3\u8868\u4e5d\u4e2d\u53ef\u770b\u51fa\u96d6\u7136\u5728\u4eba\u5de5\u8fa8\u8b58\u4e0a\u53ef\u5f97\u5230\u81f3\u5c11 60%\u7684\u8fa8\u8b58\u7387\uff0c\u4f46\u5728\u57fa\u6e96\u5be6\u9a57 \u4e2d\u8a31\u591a\u60c5\u7dd2\u7684\u8fa8\u8b58\u7387\u4e0d\u53ca 60%\u3002\u6b64\u5be6\u9a57\u7d50\u679c\u4e5f\u53cd\u6620\u51fa\u9304\u97f3\u72c0\u6cc1\uff0c\u4e03\u7a2e\u60c5\u7dd2\u4e2d\u4ee5\u751f\u6c23\u3001\u7121 \u7b54\u6848 \u751f\u6c23 \u7121\u804a \u5641\u5fc3 \u5bb3\u6015 \u958b\u5fc3 \u4e2d\u6027 \u50b7\u5fc3 \u6e96\u78ba\u7387 \u7d50\u679c 13 \u804a\u548c\u5bb3\u6015\u4e09\u7a2e\u60c5\u7dd2\u7684\u8fa8\u8b58\u7387\u8f03\u4f73\uff0c\u6b64\u4e09\u7a2e\u60c5\u7dd2\u5c0d\u9304\u97f3\u8005\u4f86\u8aaa\u8f03\u597d\u8868\u9054\uff1b\u800c\u9304\u97f3\u8005\u666e\u904d\u8a8d \u751f\u6c23 44 1 2 0 6 1 4 72.6 4.1 \u5be6\u9a57\u8a2d\u5b9a \u6578\u7e2e\u653e\u70ba 0 \u81f3 1\uff0c\u516c\u5f0f\u5b9a\u7fa9\u70ba\u7b97\u5f0f(1)\uff0cV \u8207 V'\u5206\u5225\u70ba\u6b63\u898f\u5316\u524d\u8207\u6b63\u898f\u5316\u5f8c\u7684\u503c\u3002\u5f8c\u7aef\u4ee5 \u652f\u6301\u5411\u91cf\u6a5f\u505a\u70ba\u5206\u985e\u5668\uff0c\u652f\u6301\u5411\u91cf\u6a5f\u7684\u6838\u5fc3\u51fd\u5f0f(kernel function)\u70ba\u57fa\u65bc\u5e8f\u5217\u6700\u5c0f\u5316\u6f14\u7b97 \u6cd5(Sequential Minimal Optimization)\u7684\u591a\u9805\u5f0f\u6838\u5fc3(Polynomial Kernel)\uff0cKernel \u51fd\u5f0f\u70ba\u7b97\u5f0f (2)\uff0cp \u8a2d\u70ba 1\u3002\u5be6\u9a57\u65b9\u5f0f\u4ee5\u6bcf\u6b21\u5f9e\u8a9e\u6599\u5eab\u4e2d\u6311\u51fa\u4e00\u4f4d\u8a9e\u8005\u4f5c\u70ba\u6e2c\u8a66\u8a9e\u8005\uff0c\u800c\u5176\u4ed6\u8a9e\u8005\u7576 \u4f5c\u8a13\u7df4\u8a9e\u8005(Leave-One-Speaker-Out, LOSO)\uff0c\u7d93\u904e 10 \u6b21\u8a13\u7df4\u8207\u8fa8\u8b58\u5f8c\u7d71\u8a08\u7d50\u679c\u4fbf\u70ba\u6b64\u8a9e \u6599\u5eab\u7684\u8fa8\u8b58\u7387\u3002 \u8868\u516d\u300139 \u500b\u6cdb\u51fd Functionals, etc. \u70ba\u5641\u5fc3\u53ea\u55ae\u7d14\u5229\u7528\u8a9e\u8abf\u4f86\u8868\u9054\u4e26\u4e0d\u5920\u76f4\u89ba\uff0c\u56e0\u6b64\u5728\u9304\u97f3\u7684\u60c5\u7dd2\u8868\u9054\u4e0a\u6703\u76f8\u5c0d\u8f03\u5dee\uff0c\u4e09\u7a2e \u7121\u804a 0 59 0 0 0 6 1 86.7 \u8a9e\u8a00\u7684\u6700\u9ad8\u8fa8\u8b58\u7387\u50c5 50%\uff1b\u53e6\u5728\u9304\u97f3\u904e\u7a0b\u4e2d\u50b7\u5fc3\u548c\u7121\u804a\u5169\u7a2e\u60c5\u7dd2\u5c0d\u8a9e\u8005\u8868\u9054\u4f86\u8aaa\u6703\u6709\u76f8 \u4f3c\u7684\u72c0\u6cc1\uff0c\u5be6\u9a57\u4e5f\u53ef\u770b\u51fa\u50b7\u5fc3\u7684\u8fa8\u8b58\u7d50\u679c\u6709\u4e0d\u5c11\u90e8\u5206\u88ab\u5206\u985e\u70ba\u7121\u804a\uff0c\u56e0\u6b64\u9020\u6210\u50b7\u5fc3\u7684\u8fa8 \u5641\u5fc3 6 4 21 0 2 3 5 50.0 # of fun. Respective rel. position of max./min. value \u8b58\u7387\u666e\u904d\u8f03\u4f4e\u3002\u5728\u53f0\u8a9e\u7684\u4e2d\u6027\u50c5 17.2%\u7684\u6e96\u78ba\u7387\uff0c\u63a8\u6e2c\u56e0\u5404\u500b\u8a9e\u8005\u5728\u8868\u9054\u65b9\u5f0f\u4e0a\u7684\u5dee\u7570 \u5bb3\u6015 2 0 0 45 6 3 6 64.1 2 Range (max.-min.) \u6027\u904e\u5927\u9020\u6210\u5206\u985e\u5668\u7121\u6cd5\u9806\u5229\u8fa8\u8b58\uff0c\u4e3b\u8981\u70ba\u8a9e\u901f\u8207\u8072\u8abf\u7684\u5dee\u7570\uff0c\u800c\u4eba\u5de5\u8fa8\u8b58\u6642\u4eba\u5011\u53ef\u4ee5\u7531 \u958b\u5fc3 7 0 4 6 16 2 8 33.3 1 Max. and min. value -arithmetic mean \u8a9e\u8005\u7684\u7279\u6027\u53bb\u63a8\u6572\u51fa\u5176\u60c5\u7dd2\uff0c\u4f46\u8fa8\u8b58\u5668\u4e0a\u537b\u7121\u6cd5\u505a\u51fa\u6b64\u5224\u65b7\uff0c\u6240\u4ee5\u9020\u6210\u8fa8\u8b58\u7387\u4e0d\u7406\u60f3\u7684 \u50b7\u5fc3 1 10 7 2 1 15 2 28.8 2 Arithmetic mean, Quadratic mean \u72c0\u6cc1\u3002 \u4e2d\u6027 5 4 4 3 5 0 39 62.9 2 Number of non-zero values 1 Geometric, and quadratic mean of non-zero values 2 \u8868\u4e03\u3001\u570b\u8a9e\u57fa\u6e96\u5be6\u9a57\u6df7\u6dc6\u77e9\u9663 \u4e94\u3001\u7d50\u8ad6 \u7b54\u6848 \u751f\u6c23 \u7121\u804a \u5641\u5fc3 \u5bb3\u6015 \u958b\u5fc3 \u4e2d\u6027 \u50b7\u5fc3 \u6e96\u78ba\u7387 \u7d50\u679c \u5728\u6b64\u7bc7\u8ad6\u6587\u4e2d\u6211\u5011\u5efa\u7acb\u4e86\u4e00\u500b\u53f0\u7063\u8a9e\u8a00\u7684\u60c5\u7dd2\u8a9e\u6599\u5eab\uff0c\u5176\u4e2d\u4ee5\u53f0\u7063\u5e38\u898b\u7684\u570b\u8a9e\u3001\u53f0 \u904e\u5e74\u611b\u8cb7\u4e00\u4ef6\u65b0\u886b t06 \u5641\u5fc3 46 55 54 155 Mean of absolute values, Mean of non-zero abs. values 2 \u751f\u6c23 44 1 2 0 8 0 3 75.9 \u8a9e\u548c\u5ba2\u5bb6\u8a9e\u4e09\u7a2e\u8a9e\u8a00\u9304\u88fd\u800c\u6210\u3002\u6211\u5011\u4eff\u7167\u516c\u958b\u8a9e\u6599\u5eab EMO-DB \u9304\u88fd\uff0c\u5171\u5305\u542b\u4e86\u751f\u6c23\u3001\u7121 \u9019\u79ae\u62dc\u653e\u5047\uff0c\u6b32\u4f6e\u670b\u53cb\u505a\u4f19\u51fa\u53bb\u4f01\u6843 t07 \u6628\u900f\u65e9\u6b32\u51fa\u9580\u7684\u6642\u9663\uff0c\u5916\u5957\u88ab\u9264\u4ed4\u9264\u5230 t08 \u4eca\u4ed4\u65e5\u898f\u5de5\u650f\u7121\u98df\u7269\u4ef6\u5c31\u592d t09 \u9019\u73ed\u706b\u8eca\u4eba\u8db3\u6fdf\uff0c\u771f\u6fdf\u4eba\u650f\u7121\u4f4d\u6e6f\u5750 t10 \u900f\u65e9\u53bb\u9a0e\u8173\u8e0f\u8eca\uff0c\u4e0b\u6661\u6563\u6b65\u53bb\u8cb7\u7269\u4ef6 \u5bb3\u6015 57 61 64 182 \u958b\u5fc3 67 58 48 Quartiles and inter-quartile ranges 6 \u7121\u804a 0 59 0 0 0 4 3 89.4 \u804a\u3001\u5641\u5fc3\u3001\u5bb3\u6015\u3001\u958b\u5fc3\u3001\u4e2d\u6027\u548c\u50b7\u5fc3\u5171\u4e03\u7a2e\uff0c\u6bcf\u7a2e\u8a9e\u8a00\u7684\u8a9e\u8005\u70ba\u5341\u4f4d\uff0c\u5206\u5225\u70ba\u4e94\u4f4d\u7537\u6027 (1) 95% and 98% percentile 2 \u5641\u5fc3 6 4 21 0 7 4 4 45.7 \u4ee5\u53ca\u4e94\u4f4d\u5973\u6027\uff0c\u6587\u672c\u5171\u5341\u53e5\uff0c\u6bcf\u4f4d\u8a9e\u8005\u4ee5\u4e03\u7a2e\u60c5\u7dd2\u4f86\u8868\u9054\u540c\u4e00\u53e5\u6587\u672c\u3002\u5728\u9304\u88fd\u5f8c\u9032\u884c\u4eba 173 \u50b7\u5fc3 57 56 52 (2) Std. deviation, variance, kurtosis, skewness 4 \u5bb3\u6015 2 0 0 45 6 2 2 78.9 \u5de5\u8fa8\u8b58\u7684\u6aa2\u9a57\uff0c\u4ee5\u4eba\u5de5\u8fa8\u8b58\u7387 60%\u4f5c\u70ba\u7be9\u9078\u57fa\u6e96\uff0c\u4ee5\u6b64\u78ba\u4fdd\u6211\u5011\u8a9e\u6599\u5177\u6709\u4e00\u5b9a\u7684\u8fa8\u5225\u5ea6\uff0c 165 \u4e2d\u6027 87 64 62 Centroid 1 \u958b\u5fc3 7 0 4 6 40 1 9 \u7be9\u9078\u5f8c\u4e09\u7a2e\u8a9e\u8a00\u7684\u6240\u5269\u53e5\u6578\u70ba\u570b\u8a9e 438 \u53e5\u3001\u53f0\u8a9e 422 \u53e5\u548c\u5ba2\u5bb6\u8a9e 401 \u53e5\u3002\u53e6\u6211\u5011\u4f7f\u7528\u4e86 59.7 213 \u7e3d\u6578 438 422 401 1261 4.2 \u57fa\u6e96\u7279\u5fb5\u96c6 \u672c\u8ad6\u6587\u4f7f\u7528\u8072\u5b78\u7279\u5fb5\u4f5c\u70ba\u5be6\u9a57\u7684\u7279\u5fb5\u53c3\u6578\uff0c\u8868\u4e94\u548c\u8868\u516d\u5206\u5225\u5217\u51fa\u5927\u578b\u7279\u5fb5\u96c6\u6240\u5305\u542b Zero-crossing rate 1 # of peaks, mean dist. btwn. peaks, arth. mean of peaks, arth. Mean of peaks -overall arth. Mean 4 \u50b7\u5fc3 1 10 7 2 3 31 3 \u4e00\u500b\u9f90\u5927\u7684\u8072\u5b78\u7279\u5fb5\u96c6\u4e26\u642d\u914d\u652f\u6301\u5411\u91cf\u6a5f\u505a\u70ba\u6211\u5011\u7684\u57fa\u6e96\u5be6\u9a57\uff0c\u5176\u4e2d\u5206\u5225\u5728\u4e09\u7a2e\u8a9e\u8a00\u53ef 54.4 \u4e2d\u6027 5 4 4 3 10 1 60 \u4ee5\u5f97\u5230\u570b\u8a9e 68.5%\u3001\u53f0\u8a9e 50.7%\u548c\u5ba2\u5bb6\u8a9e 58.5%\u7684\u8fa8\u8b58\u7387\uff0c\u4fbf\u70ba\u4e09\u500b\u8a9e\u8a00\u7684\u57fa\u6e96\u8fa8\u8b58\u7387\u3002 69.0 \u6b64\u8cc7\u6599\u5eab\u53ca\u5be6\u9a57\u6578\u64da\u53ef\u505a\u70ba\u672a\u4f86\u53f0\u7063\u8a9e\u97f3\u60c5\u7dd2\u8fa8\u8b58\u4f7f\u7528\u53ca\u53c3\u8003\u3002\u5728\u672a\u4f86\u82e5\u53ef\u4ee5\u6211\u5011\u5e0c\u671b \u7684\u4f4e\u968e\u53c3\u6578(low-level descriptor, LLD)\u4ee5\u53ca\u6cdb\u51fd(functional)\u3002\u9996\u5148\u5f9e\u97f3\u8a0a\u4e2d\u64f7\u53d6\u6bcf\u500b\u97f3\u6846 Linear regression coefficients and corresp. approximation error 4 \u80fd\u5920\u66f4\u63d0\u5347\u8a9e\u6599\u5eab\u7684\u54c1\u8cea\uff0c\u5982\u6539\u5584\u53f0\u8a9e\u7684\u4e2d\u6027\u8fa8\u8b58\u7387\u904e\u4f4e\u7684\u72c0\u6cc1\uff0c\u5c0d\u65bc\u7be9\u9078\u5f8c\u53e5\u6578\u5269\u9918 \u70ba 56 \u7dad\u7684\u7279\u5fb5\uff0c\u518d\u900f\u904e\u6cdb\u51fd\u5c07\u9019\u4e9b\u97f3\u6846\u7279\u5fb5\u8f49\u63db\u70ba\u4e00\u500b\u53e5\u5b50\u4e00\u7d44\u7684\u7279\u5fb5\u5411\u91cf\uff0c56 \u500b\u4f4e \u968e\u53c3\u6578\u548c 39 \u500b\u6cdb\u51fd\u518d\u8a08\u7b97\u4e00\u968e\u8207\u4e8c\u968e\u52d5\u614b\u7279\u5fb5\u5f8c\u5171\u5f97\u5230 6552 \u500b\u7279\u5fb5\u53c3\u6578\uff0c\u6b64\u4fbf\u70ba\u6211\u5011 Quadratic regression coefficients and corresp. approximation error 5 \u904e\u5c11\u7684\u8a9e\u8005\u9032\u884c\u91cd\u9304\u6216\u4ee5\u65b0\u7684\u8a9e\u8005\u8cc7\u6599\u4ee3\u66ff\uff0c\u63d0\u5347\u54c1\u8cea\u7684\u540c\u6642\u4e5f\u589e\u52a0\u8a9e\u6599\u5eab\u7684\u7e3d\u53e5\u6578\u3002</td></tr><tr><td>\u9a57\uff0c\u6700\u5f8c\u70ba\u6211\u5011\u7684\u7d50\u8ad6\u3002</td></tr></table>" | |
| } | |
| } | |
| } | |
| } |