File size: 72,550 Bytes
6fa4bc9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
{
    "paper_id": "O04-1025",
    "header": {
        "generated_with": "S2ORC 1.0.0",
        "date_generated": "2023-01-19T07:59:51.258047Z"
    },
    "title": "",
    "authors": [
        {
            "first": "",
            "middle": [],
            "last": "\u9673\u6c5f\u6751",
            "suffix": "",
            "affiliation": {},
            "email": ""
        },
        {
            "first": "",
            "middle": [],
            "last": "\u7f85\u745e\u9e9f",
            "suffix": "",
            "affiliation": {},
            "email": ""
        },
        {
            "first": "",
            "middle": [],
            "last": "\u5f35\u667a\u661f",
            "suffix": "",
            "affiliation": {},
            "email": ""
        },
        {
            "first": "",
            "middle": [],
            "last": "\u674e\u4fca\u4ec1",
            "suffix": "",
            "affiliation": {},
            "email": ""
        },
        {
            "first": "\u8cc7\u8a0a\u5de5\u7a0b\u7cfb",
            "middle": [],
            "last": "\u570b\u7acb\u6e05\u83ef\u5927\u5b78",
            "suffix": "",
            "affiliation": {},
            "email": ""
        },
        {
            "first": "",
            "middle": [],
            "last": "\u65b0\u7af9\u5e02\u5149\u5fa9\u8def\u4e8c\u6bb5",
            "suffix": "",
            "affiliation": {},
            "email": ""
        },
        {
            "first": "\u4e2d\u83ef\u96fb\u4fe1\u7814\u7a76\u6240",
            "middle": [],
            "last": "\u6843\u5712\u7e23\u694a\u6885\u93ae\u6c11\u65cf\u8def",
            "suffix": "",
            "affiliation": {},
            "email": ""
        }
    ],
    "year": "",
    "venue": null,
    "identifiers": {},
    "abstract": "",
    "pdf_parse": {
        "paper_id": "O04-1025",
        "_pdf_hash": "",
        "abstract": [],
        "body_text": [
            {
                "text": "Z D d IH I OW o TH L AE G DH F IY i OY W UH U AH O EH A JH P P p UW u AO R ER S K k R r V v AW aU EY e L l S s W w AY aI F f M m SH B Y j B b G g N n SIL sil Z z CH Q HH h NG E T t ZH N \u6211\u5011\u4f7f\u7528\u4ee5\u4e0b\u4e09\u7a2e\u539f\u5247\u4f86\u5c0d TIMIT \u7684",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            }
        ],
        "back_matter": [],
        "bib_entries": {
            "BIBREF0": {
                "ref_id": "b0",
                "title": "\uff0c\u5317\u4eac\u6e05\u83ef\u5927\u5b78\u78a9\u58eb\u8ad6\u6587\uff0c\u6c11\u570b 91 \u5e74",
                "authors": [
                    {
                        "first": "",
                        "middle": [],
                        "last": "\u9418\u6797\uff0c\"\u6f22\u8a9e\u8a9e\u97f3\u8fa8\u5225\u8aaa\u8a71\u9a57\u8b49",
                        "suffix": ""
                    }
                ],
                "year": null,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "\u9418\u6797\uff0c\"\u6f22\u8a9e\u8a9e\u97f3\u8fa8\u5225\u8aaa\u8a71\u9a57\u8b49\"\uff0c\u5317\u4eac\u6e05\u83ef\u5927\u5b78\u78a9\u58eb\u8ad6\u6587\uff0c\u6c11\u570b 91 \u5e74",
                "links": null
            },
            "BIBREF2": {
                "ref_id": "b2",
                "title": "A probabilistic approach to AMDF pitch detection, Spoken Language",
                "authors": [
                    {
                        "first": "G",
                        "middle": [
                            "S"
                        ],
                        "last": "Ying",
                        "suffix": ""
                    },
                    {
                        "first": "L",
                        "middle": [
                            "H"
                        ],
                        "last": "Jamieson",
                        "suffix": ""
                    },
                    {
                        "first": "C",
                        "middle": [
                            "D"
                        ],
                        "last": "Michell",
                        "suffix": ""
                    }
                ],
                "year": 1996,
                "venue": "ICSLP 96. Proceedings",
                "volume": "2",
                "issue": "",
                "pages": "1201--1204",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "G.S. Ying, L.H. Jamieson and C.D. Michell, A probabilistic approach to AMDF pitch detection, Spoken Language, 1996. ICSLP 96. Proceedings., Fourth International Conference on Volume: 2 , 1996 , Page(s): 1201-1204 vol.2",
                "links": null
            },
            "BIBREF3": {
                "ref_id": "b3",
                "title": "The HTK Book version 3",
                "authors": [
                    {
                        "first": "Steve",
                        "middle": [],
                        "last": "Young",
                        "suffix": ""
                    }
                ],
                "year": 2000,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Steve Young, The HTK Book version 3, Microsoft Corporation, 2000",
                "links": null
            },
            "BIBREF4": {
                "ref_id": "b4",
                "title": "Fundamentals of speech recognition",
                "authors": [
                    {
                        "first": "Lawrence",
                        "middle": [],
                        "last": "Rabiner",
                        "suffix": ""
                    },
                    {
                        "first": "B",
                        "middle": [],
                        "last": "Juang",
                        "suffix": ""
                    }
                ],
                "year": 1993,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Lawrence Rabiner, B.H Juang, Fundamentals of speech recognition, Prentice Hall, 1993",
                "links": null
            },
            "BIBREF5": {
                "ref_id": "b5",
                "title": "Discrete-Time Processing of Speech Signals",
                "authors": [
                    {
                        "first": "J",
                        "middle": [
                            "D"
                        ],
                        "last": "",
                        "suffix": ""
                    },
                    {
                        "first": "J",
                        "middle": [
                            "G"
                        ],
                        "last": "",
                        "suffix": ""
                    },
                    {
                        "first": "J",
                        "middle": [
                            "H"
                        ],
                        "last": "",
                        "suffix": ""
                    },
                    {
                        "first": "L",
                        "middle": [
                            "H"
                        ],
                        "last": "",
                        "suffix": ""
                    }
                ],
                "year": 1993,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "J.D., J.G., J.H. and L.H., Discrete-Time Processing of Speech Signals, Prentice Hall, 1993",
                "links": null
            },
            "BIBREF6": {
                "ref_id": "b6",
                "title": "Mnophonic transcription with autocorrelation",
                "authors": [
                    {
                        "first": "Giuliano",
                        "middle": [],
                        "last": "Monti",
                        "suffix": ""
                    },
                    {
                        "first": "Mark",
                        "middle": [],
                        "last": "Sandler",
                        "suffix": ""
                    }
                ],
                "year": 2000,
                "venue": "Proceedings of the COST G-6 Conference on Digital Audio Effects (DAFX-00)",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Giuliano Monti, Mark Sandler, Mnophonic transcription with autocorrelation, Proceedings of the COST G-6 Conference on Digital Audio Effects (DAFX-00), Verona, Italy, December 7-9, 2000",
                "links": null
            },
            "BIBREF7": {
                "ref_id": "b7",
                "title": "Automatic scoring of pronunciation quality",
                "authors": [
                    {
                        "first": "L",
                        "middle": [],
                        "last": "Neumeyer",
                        "suffix": ""
                    },
                    {
                        "first": "H",
                        "middle": [],
                        "last": "Franco",
                        "suffix": ""
                    },
                    {
                        "first": "V",
                        "middle": [],
                        "last": "Digalakis",
                        "suffix": ""
                    },
                    {
                        "first": "M",
                        "middle": [],
                        "last": "Weintraub",
                        "suffix": ""
                    }
                ],
                "year": 1999,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "L. Neumeyer, H. Franco, V. Digalakis and M. Weintraub, Automatic scoring of pronunciation quality, 1999",
                "links": null
            },
            "BIBREF8": {
                "ref_id": "b8",
                "title": "Automatic pronunciation scoring for language instruction",
                "authors": [
                    {
                        "first": "H",
                        "middle": [],
                        "last": "Franco",
                        "suffix": ""
                    },
                    {
                        "first": "L",
                        "middle": [],
                        "last": "Neumeyer",
                        "suffix": ""
                    },
                    {
                        "first": "Y",
                        "middle": [],
                        "last": "Kim",
                        "suffix": ""
                    },
                    {
                        "first": "O",
                        "middle": [],
                        "last": "Ronen",
                        "suffix": ""
                    }
                ],
                "year": 1997,
                "venue": "Proc. Int. Congress on Acoustics, Speech and Signal Processing",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "H. Franco, L. Neumeyer, Y. Kim and O. Ronen, Automatic pronunciation scoring for language instruction, Proc. Int. Congress on Acoustics, Speech and Signal Processing(ICASSP), 1997",
                "links": null
            },
            "BIBREF9": {
                "ref_id": "b9",
                "title": "Neuro-Fuzzy and Soft Computing",
                "authors": [
                    {
                        "first": "J.-S",
                        "middle": [],
                        "last": "Roger",
                        "suffix": ""
                    },
                    {
                        "first": "C.-T",
                        "middle": [],
                        "last": "Jang",
                        "suffix": ""
                    },
                    {
                        "first": "E",
                        "middle": [],
                        "last": "Sun",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Mizutani",
                        "suffix": ""
                    }
                ],
                "year": 1996,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "J.-S. Roger. Jang, C.-T. Sun, and E. Mizutani, Neuro-Fuzzy and Soft Computing, Prentice Hall, 1996",
                "links": null
            },
            "BIBREF10": {
                "ref_id": "b10",
                "title": "Pattern Recognition Principles",
                "authors": [
                    {
                        "first": "J",
                        "middle": [
                            "T"
                        ],
                        "last": "Tou",
                        "suffix": ""
                    },
                    {
                        "first": "R",
                        "middle": [
                            "C"
                        ],
                        "last": "Gonzalez",
                        "suffix": ""
                    }
                ],
                "year": 1974,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "J. T. Tou and R. C. Gonzalez, Pattern Recognition Principles, Addison-Wesley Publishing Company, 1974",
                "links": null
            },
            "BIBREF11": {
                "ref_id": "b11",
                "title": "\uff0c\u6e05\u83ef\u5927\u5b78\u78a9\u58eb\u8ad6\u6587\uff0c\u6c11\u570b 91 \u5e74",
                "authors": [
                    {
                        "first": "",
                        "middle": [],
                        "last": "\u674e\u4fca\u6bc5\uff0c\"\u8a9e\u97f3\u8a55\u5206",
                        "suffix": ""
                    }
                ],
                "year": null,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "\u674e\u4fca\u6bc5\uff0c\"\u8a9e\u97f3\u8a55\u5206\"\uff0c\u6e05\u83ef\u5927\u5b78\u78a9\u58eb\u8ad6\u6587\uff0c\u6c11\u570b 91 \u5e74",
                "links": null
            },
            "BIBREF12": {
                "ref_id": "b12",
                "title": "Utterance Verification based on the Likelihood Distance to Alternative Paths",
                "authors": [
                    {
                        "first": "Gies",
                        "middle": [],
                        "last": "Bouwman",
                        "suffix": ""
                    },
                    {
                        "first": "Lou",
                        "middle": [],
                        "last": "Boves",
                        "suffix": ""
                    }
                ],
                "year": 2002,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Gies Bouwman and Lou Boves, Utterance Verification based on the Likelihood Distance to Alternative Paths, Department of Speech, University of Nijmegen, The Netherlands, 2002",
                "links": null
            },
            "BIBREF13": {
                "ref_id": "b13",
                "title": "Vocabulary Independent Discriminative Utterance Verification for Nonkeyword Rejection in Subword based Speech Recognition",
                "authors": [
                    {
                        "first": "A",
                        "middle": [],
                        "last": "Rafid",
                        "suffix": ""
                    },
                    {
                        "first": "Chin-Hui",
                        "middle": [],
                        "last": "Sukkar",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Lee",
                        "suffix": ""
                    }
                ],
                "year": 1996,
                "venue": "IEEE Transactions on Speech and Audio Processing",
                "volume": "4",
                "issue": "6",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Rafid A. Sukkar and Chin-Hui Lee, Vocabulary Independent Discriminative Utterance Verification for Nonkeyword Rejection in Subword based Speech Recognition, IEEE Transactions on Speech and Audio Processing, VOL. 4, No. 6, November 1996",
                "links": null
            },
            "BIBREF14": {
                "ref_id": "b14",
                "title": "Automatic Text-Independent Pronunciation Scoring of Foreign Language Student Speech",
                "authors": [
                    {
                        "first": "Leonardo",
                        "middle": [],
                        "last": "Neumeyer",
                        "suffix": ""
                    },
                    {
                        "first": "Horacio",
                        "middle": [],
                        "last": "Franco",
                        "suffix": ""
                    },
                    {
                        "first": "Mitchel",
                        "middle": [],
                        "last": "Weintraub",
                        "suffix": ""
                    },
                    {
                        "first": "Patti",
                        "middle": [],
                        "last": "Price",
                        "suffix": ""
                    }
                ],
                "year": 1996,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Leonardo Neumeyer, Horacio Franco, Mitchel Weintraub, and Patti Price, Automatic Text-Independent Pronunciation Scoring of Foreign Language Student Speech, 1996",
                "links": null
            },
            "BIBREF15": {
                "ref_id": "b15",
                "title": "Automatic Evaluation of Dutch Pronunciation by Using Speech Recognition Technology",
                "authors": [
                    {
                        "first": "C",
                        "middle": [],
                        "last": "Cucchiarini",
                        "suffix": ""
                    },
                    {
                        "first": "H",
                        "middle": [],
                        "last": "Strik",
                        "suffix": ""
                    },
                    {
                        "first": "L",
                        "middle": [],
                        "last": "Boves",
                        "suffix": ""
                    }
                ],
                "year": 1997,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "C. Cucchiarini, H. Strik and L. Boves, Automatic Evaluation of Dutch Pronunciation by Using Speech Recognition Technology, Department of Speech, University of Nijmegen, The Netherlands, 1997",
                "links": null
            },
            "BIBREF16": {
                "ref_id": "b16",
                "title": "Automatic Pronunciation Assessment for Mandarin Chinese",
                "authors": [
                    {
                        "first": "Jyh-Shing Roger",
                        "middle": [],
                        "last": "Jiang-Chun Chen",
                        "suffix": ""
                    },
                    {
                        "first": "Jun-Yi",
                        "middle": [],
                        "last": "Jang",
                        "suffix": ""
                    },
                    {
                        "first": "Ming-Chun",
                        "middle": [],
                        "last": "Li",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Wu",
                        "suffix": ""
                    }
                ],
                "year": 2004,
                "venue": "Proc. Int. Conf. on Multimedia And Expo (ICME)",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Jiang-Chun Chen, Jyh-Shing Roger Jang, Jun-Yi Li and Ming-Chun Wu, \"Automatic Pronunciation Assessment for Mandarin Chinese\", Proc. Int. Conf. on Multimedia And Expo (ICME), 2004",
                "links": null
            }
        },
        "ref_entries": {
            "TABREF0": {
                "content": "<table><tr><td>\u7684\u8a55\u5206\u6a19\u6e96[20]\u3002 \u5207\u5272\u51fa\u4f86\u7684\u97f3\u7d20\u5f88\u5b8c\u6574\uff0c\u800c\u4e0b\u65b9\u7684\u8a9e\u97f3\u5167\u5bb9\u548c\u6a19\u6e96\u8a9e\u97f3\u5167\u5bb9\u4e0d\u76e1\u76f8\u540c\uff0c\u56e0\u800c\u8fa8\u8b58\u7a0b\u5f0f\u5c07\u6a39\u72c0\u7db2\u8def\u5c55\u958b\u81f3\u7bc0\u9ede \u7d20\uff0c\u53e6\u4e00\u7a2e\u5247\u662f\u8a9e\u97f3\u8a0a\u865f\u7684\u5f8c\u534a\u90e8\u53ef\u80fd\u6c92\u6709\u8fa6\u6cd5\u5207\u5272\u51fa\u97f3\u7d20\u3002\u800c\u5728\u9019\u4e00\u7bc0\u8a0e\u8ad6\u7684\u9a57\u8b49\u6a5f\u5236\uff0c\u4e3b\u8981\u662f\u91dd\u5c0d\u524d\u8005\u7684 \u8a71\u4e2d\u82e5\u5b58\u5728\u9023\u7e8c 3 \u500b\u55ae\u5b57\u4ee5\u4e0a\u548c\u6a19\u6e96\u8a9e\u97f3\u5167\u5bb9\u76f8\u540c\uff0c\u4f46\u4e26\u4e0d\u662f\u5b8c\u5168\u76f8\u540c\uff0c\u5373\u70ba\u300c\u90e8\u4efd\u76f8\u540c\u300d\u3002</td></tr><tr><td>1 \u524d\u8a00 \u7531\u65bc\u8fd1\u5e74\u4f86\u96fb\u8166\u8a08\u7b97\u80fd\u529b\u7684\u63d0\u6607\u4ee5\u53ca\u8a9e\u97f3\u8fa8\u8b58\u6280\u8853\u7684\u9032\u6b65\uff0c\u8a9e\u97f3\u8655\u7406\u5728\u6211\u5011\u65e5\u5e38\u751f\u6d3b\u4e0a\u7684\u61c9\u7528\u8207\u65e5\u4ff1\u589e\uff0c\u5982\u8a9e \u97f3\u8fa8\u8b58\u3001\u8a9e\u97f3\u5408\u6210\u3001\u8a9e\u8005\u8b58\u5225\u7b49\u7b49\u3002\u5176\u4e2d\uff0c\u5728\u8de8\u570b\u754c\u7684\u8a9e\u8a00\u5b78\u7fd2\u4e2d\uff0c\u4ee5\u96fb\u8166\u8f14\u52a9\u4f7f\u7528\u8005\u9032\u884c\u975e\u6bcd\u8a9e\u5b78\u7fd2(CALL, Segment Duration[10]\u3002 \u81f3\u65bc\u82f1\u6587\u7684\u8a9e\u97f3\u8a55\u5206\uff0c2002 \u5e74\u6e05\u83ef\u5927\u5b78\u7684\u674e\u4fca\u6bc5\u4ee5\u6885\u723e\u5012\u983b\u8b5c\u3001Magnitude \u53ca Pitch \u4e09\u7a2e\u8a55\u5206\u53c3\u6578\u89c0\u5bdf\u5c0d \u82f1\u6587\u8a9e\u97f3\u8a55\u5206\u7684\u5f71\u97ff\uff0c\u5176\u5be6\u9a57\u767c\u73fe\u6885\u723e\u5012\u983b\u8b5c\u53c3\u6578\u5c0d\u82f1\u6587\u8a9e\u97f3\u8a55\u5206\u7684\u91cd\u8981\u6027\u6700\u5927\uff0c\u53e6\u5916\u4ed6\u4e5f\u5c07\u5404\u500b\u7279\u5fb5\u7684\u5dee\u7570 \u7a0b\u5ea6\u8f49\u63db\u6210\u5206\u6578\uff0c\u4ee5\u56de\u994b\u7d66\u4f7f\u7528\u8005\u53c3\u8003[15]\u30022004 \u5e74\u9673\u6c5f\u6751\u548c\u5f35\u667a\u661f\u7b49\u4eba\u5229\u7528\u4e86 HMM \u548c GMM \u5206\u5225\u5c0d\u4e2d\u6587\u7684 \u767c\u97f3\u548c\u8072\u8abf\u9032\u884c\u8a55\u5206\uff0c\u4e26\u4ee5 Downhill Simplex Search \u9032\u884c\u4e86\u8a55\u5206\u7cfb\u7d71\u53c3\u6578\u7684\u6700\u4f73\u5316\uff0c\u4ee5\u6c42\u9054\u5230\u548c\u4e2d\u6587\u5c08\u5bb6\u4e00\u81f4 \u63a5\u4e0b\u4f86\u7684\u8ad6\u8ff0\u4e2d\uff0c\u9996\u5148\u6211\u5011\u63d0\u51fa\u5be6\u4f5c\u300c\u8aaa\u8a71\u9a57\u8b49\u300d\u7684\u65b9\u6cd5\uff0c\u5305\u542b\u8072\u5b78\u6a21\u578b\u76f8\u4f3c\u5ea6\u6392\u540d\u3001\u9a57\u8b49\u7cfb\u7d71\u7684\u5efa\u7acb \u53ca\u9a57\u8b49\u7cfb\u7d71\u7684\u53ef\u9760\u6027\u7b49\u3002\u63a5\u8457\u662f\u300c\u8a9e\u97f3\u8a0a\u865f\u5207\u5272\u300d\u3002\u9019\u90e8\u4efd\u5305\u542b\u96b1\u85cf\u5f0f\u99ac\u53ef\u592b\u6a21\u578b(Hidden Markov Model)\u7684\u8a13 \u7df4\u548c\u4ee5\u7dad\u7279\u6bd4\u6f14\u7b97\u6cd5(Viterbi algorithm)\u70ba\u57fa\u790e\u7684\u8a9e\u97f3\u8a0a\u865f\u5207\u5272\u6280\u5de7\u3002\u518d\u4f86\u662f\u300c\u82f1\u6587\u8a9e\u97f3\u8a55\u5206\u300d\uff0c\u5176\u4e2d\u63d0\u5230\u4e86\u95dc \u65bc\u8a55\u5206\u53c3\u6578\u7684\u64f7\u53d6\u3001\u8a55\u5206\u53c3\u6578\u6b63\u898f\u5316\u3001\u5716\u6a23\u6bd4\u5c0d\u6d41\u7a0b\u3001\u8a55\u5206\u6a5f\u5236\u7684\u5efa\u7acb\u7b49\uff0c\u4e26\u8a2d\u8a08\u5be6\u9a57\u4ee5\u6c42\u51fa\u5404\u8a55\u5206\u53c3\u6578\u5728\u82f1 \u6587\u8a9e\u97f3\u8a55\u5206\u4e2d\u7684\u6b0a\u91cd\uff0c\u4ee5\u7b26\u5408\u4eba\u985e\u5c08\u5bb6\u5c0d\u82f1\u6587\u8a9e\u53e5\u597d\u58de\u7684\u770b\u6cd5\u3002\u6700\u5f8c\u662f\u7e3d\u7d50\u53ca\u4eca\u5f8c\u7814\u7a76\u5de5\u4f5c\u7684\u5c55\u671b\u3002 3 \u82f1\u8a9e\u8a55\u5206\u7cfb\u7d71\u67b6\u69cb \u5728\u6b64\u82f1\u6587\u8a9e\u97f3\u8a55\u5206\u7cfb\u7d71\u4e2d\uff0c\u9996\u5148\u4ee5\u8aaa\u8a71\u9a57\u8b49\u505a\u70ba\u7b2c\u4e00\u9053\u6aa2\u8996\u95dc\u5361\uff0c\u723e\u5f8c\u4ee5\u8072\u5b78\u6a21\u578b\u4f86\u5c0d\u6a19\u6e96\u8a9e\u97f3\u53ca\u8a55\u5206\u8a9e\u97f3\u5207 \u5272\u97f3\u7d20\u7684\u6642\u9593\u5340\u6bb5\uff0c\u518d\u5c07\u9019\u4e9b\u8cc7\u8a0a\u9001\u81f3\u82f1\u6587\u8a9e\u97f3\u8a55\u5206\u7cfb\u7d71\u7684\u6838\u5fc3\uff0c\u5229\u7528\u5404\u7a2e\u8a55\u5206\u53c3\u6578\uff0c\u9010\u97f3\u7d20\u5730\u6bd4\u8f03\u8a55\u5206\u8a9e\u97f3 \u548c\u6a19\u6e96\u8a9e\u97f3\u7684\u5dee\u7570\u7a0b\u5ea6\u3002\u672c\u6587\u6240\u63d0\u4e4b\u82f1\u6587\u8a9e\u97f3\u8a55\u5206\u7cfb\u7d71\u67b6\u69cb\u6d41\u7a0b\uff0c\u5982\u5716\u8868 1\u6240\u793a\u3002 \u5716\u8868 1 \u82f1\u6587\u8a9e\u97f3\u8a55\u5206\u7cfb\u7d71\u6d41\u7a0b\u5716 3.1 \u8aaa\u8a71\u9a57\u8a3c \u6240\u8b02\u7684\u8aaa\u8a71\u9a57\u8b49(Utterance Verification)\uff0c\u5c31\u662f\u6211\u5011\u53ef\u4ee5\u91dd\u5c0d\u4e0d\u540c\u7684\u8a55\u5206\u8a9e\u97f3\u7522\u751f\u5224\u65b7\u6578\u503c\uff0c\u4e26\u4f9d\u6b64\u800c\u5c0d\u8a72\u8a55\u5206 \u8a9e\u97f3\u5167\u5bb9\u7684\u6b63\u78ba\u6027\u505a\u51fa\u5224\u65b7[1]\u3002\u6b64\u8aaa\u8a71\u9a57\u8b49\u6d41\u7a0b\u5982\u5716\u8868 2 \u6240\u793a\uff0c\u7576\u9a57\u8b49\u7cfb\u7d71\u63a5\u6536\u5230\u8a9e\u97f3\u8a0a\u865f\u5f8c\uff0c\u5206\u5225\u5c0d\u6bcf\u500b \u97f3\u7d20\u9032\u884c\u8a9e\u97f3\u8fa8\u8b58\uff0c\u4e4b\u5f8c\u518d\u4f9d\u8fa8\u8b58\u7d50\u679c\u7684\u6a5f\u7387\u503c\u6392\u540d\u4e26\u914d\u5408\u9a57\u8b49\u6a5f\u5236\u7d66\u4e88\u6700\u5f8c\u7684\u53ef\u4fe1\u5ea6\u503c\u3002 \u5716\u8868 2 \u8aaa\u8a71\u9a57\u8b49\u7cfb\u7d71\u6d41\u7a0b\u5716 3.1.1 \u97f3\u7d20\u5207\u5272 \u9019\u88e1\u5207\u5272\u7528\u7684\u6280\u8853\uff0c\u4e26\u4e0d\u662f\u7528 Viterbi Decoding \u4e2d\u5e38\u898b\u7684 Forced Alignment\uff0c\u800c\u662f\u4f7f\u7528 beam search \u4e2d pruning \u97f3\u76f8\u7576\u985e\u4f3c\uff0c\u5247\u7d93\u7531\u5207\u5272\u5f8c\u7522\u751f\u97f3\u7d20\u7684\u6578\u91cf\u5c07\u63a5\u8fd1\u751a\u81f3\u7b49\u540c\u65bc\u6a19\u6e96\u8a9e\u97f3\u97f3\u7d20\u7684\u6578\u91cf\u3002\u76f8\u53cd\u5730\uff0c\u82e5\u4e82\u8b1b\u7684\u8a55\u5206\u8a9e \u97f3\u4e2d\u53ea\u6709\u524d n \u500b\u97f3\u7d20\u548c\u6a19\u6e96\u8a9e\u97f3\u76f8\u540c(\u5f8c\u5e7e\u500b\u97f3\u7d20\u5b8c\u5168\u4e0d\u540c)\uff0c\u5247\u7d93\u7531 pruning \u5f8c\u7684\u97f3\u7d20\u4e5f\u5927\u7d04\u7b49\u65bc n\u3002\u8209\u4f8b\u4f86\u8aaa\uff0c \u5982\u679c\u6a19\u6e96\u8a9e\u97f3\u70ba\u300cshe has your dark suit in greasy wash water all year\u300d\u3001\u8a55\u5206\u8a9e\u97f3\u70ba\u300cshe has your dark suit\u300d\uff0c\u5247 \u5c0d\u65bc\u6c92\u6709\u5207\u5272\u51fa\u4f86\u7684\u97f3\u7d20\uff0c\u6211\u5011\u5247\u5c07\u5176\u53ef\u4fe1\u5ea6\u503c\u8a2d\u70ba 0\uff0c\u5982\u6b64\u4e00\u4f86\u53ef\u4ee5\u589e\u52a0\u9a57\u8b49\u7cfb\u7d71\u7684\u5340\u5225\u6027\uff0c\u4f7f\u5f97\u548c\u6a19 \u5716\u8868 3 \u70ba\u5169\u500b\u8a9e\u97f3\u7d93\u7531\u8a9e\u97f3\u8a0a\u865f\u5207\u5272\u5f8c\u7522\u751f\u7684\u4e0d\u540c\u7d50\u679c\u3002\u4e0a\u534a\u90e8\u7684\u8a9e\u97f3\u5167\u5bb9\u7b49\u540c\u65bc\u6a19\u6e96\u8a9e\u97f3\u5167\u5bb9\uff0c\u56e0\u6b64 \u7d93\u7531\u8a9e\u97f3\u8a0a\u865f\u5207\u5272\u4e4b\u5f8c\uff0c\u7522\u751f\u7684\u7d50\u679c\u53ef\u80fd\u6709\u5169\u7a2e\u60c5\u6cc1\uff1a\u4e00\u7a2e\u662f\u90e8\u4efd\u7684\u8a9e\u97f3\u8a0a\u865f\u5df2\u7d93\u6210\u529f\u5207\u5272\u51fa\u6642\u9593\u5340\u6bb5\u7684\u97f3 \u4efd\u5167\u5bb9\u548c\u6a19\u6e96\u8a9e\u97f3\u5b8c\u5168\u4e0d\u76f8\u540c\u3002\u53e6\u4e00\u90e8\u4efd\u5247\u662f\u8a9e\u97f3\u8a0a\u865f\u5167\u5bb9\u300c\u90e8\u4efd\u76f8\u540c\u300d\u65bc\u6a19\u6e96\u8a9e\u97f3\u5167\u5bb9\u3002\u5728\u6b64\u6211\u5011\u5b9a\u7fa9\u4e00\u53e5 \u6e96\u8a9e\u97f3\u5167\u5bb9\u5b8c\u5168\u4e0d\u76f8\u540c\u7684\u8a55\u5206\u8a9e\u97f3\uff0c\u5176\u53ef\u4fe1\u5ea6\u503c\u8b8a\u5f97\u76f8\u7576\u4f4e\u3002 2. Incorrect\uff1a\u53d6 168 \u53e5\u5167\u5bb9\u4e0d\u7b49\u65bc\u6a19\u6e96\u8a9e\u97f3\u5167\u5bb9\u7684\u8a9e\u6599\uff0c\u9019\u90e8\u4efd\u8a9e\u97f3\u6a94\u6848\u9577\u5ea6\u7e3d\u548c\u7d04\u70ba 7 \u5206 31 \u79d2\u3002\u5176\u4e2d\u4e00\u90e8 3.1.3 \u9a57\u8b49\u6a5f\u5236 1. Correct\uff1a \u53d6 168 \u53e5\u8aaa\u8a71\u5167\u5bb9\u76f8\u540c\u7684\u8a9e\u97f3\u8a0a\u865f\u7576\u4f5c\u6a19\u6e96\u8a9e\u97f3\u5167\u5bb9\uff0c\u9019\u90e8\u4efd\u8a9e\u97f3\u6a94\u6848\u9577\u5ea6\u7e3d\u548c\u7d04\u70ba 9 \u5206 10 \u79d2\u3002 \u7d93\u7531\u8a9e\u97f3\u8fa8\u8b58\u5f8c\uff0c\u5728\u8a55\u5206\u8a9e\u97f3\u4e2d\u6240\u80fd\u5207\u5272\u51fa\u4f86\u7684\u97f3\u7d20\u6578\u91cf\u662f 15\uff0c\u5982\u5716\u8868 3\u3002 \u540c\u7684\u97f3\u7d20\uff0c\u5373\u4f7f\u6392\u540d\u540c\u6a23\u662f\u7b2c\u4e8c\u540d\uff0c\u53ef\u662f\u548c\u7b2c\u4e00\u540d\u7684\u5c0d\u6578\u6a5f\u7387\u5dee\u8ddd\u537b\u4e0d\u76f8\u540c\uff0c\u6703\u9020\u6210\u9019\u6a23\u7684\u539f\u56e0\u5728\u65bc\u6709\u4e9b\u97f3\u7d20 \u8a55\u5206\u52d5\u4f5c\u3002\u5c0d\u6b64\u6211\u5011\u8490\u96c6\u5169\u90e8\u4efd\u7684\u5be6\u9a57\u8a9e\u6599\uff1a \u4e2d\uff0c\u53ea\u6709\u4e00\u500b model \u7684\u767c\u97f3\u548c\u8a72\u97f3\u7d20\u63a5\u8fd1\uff0c\u56e0\u6b64\u66f4\u52a0\u7a81\u986f\u4e86\u5176\u7b2c\u4e00\u548c\u7b2c\u4e8c\u540d\u7684\u5c0d\u6578\u6a5f\u7387\u5dee\u8ddd\u3002 \u884c\u8a55\u5206\u3002\u76f8\u53cd\u7684\uff0c\u5247\u8868\u793a\u9019\u53e5\u8a71\u548c\u6a19\u6e96\u8a9e\u97f3\u7684\u5167\u5bb9\u4e0d\u76f8\u540c\uff0c\u56e0\u6b64\u6211\u5011\u4e5f\u5c31\u505c\u6b62\u8b93\u5169\u53e5\u4e0d\u76f8\u540c\u7684\u8a9e\u97f3\u9032\u884c\u5f8c\u7e8c\u7684 \u540d phone model \u7684\u767c\u97f3\u5f88\u63a5\u8fd1\uff0c\u9020\u6210\u5c0d\u6578\u6a5f\u7387\u7684\u5dee\u8ddd\u76f8\u7576\u5c0f\u3002 \u800c\u5728\u4e0b\u65b9\u5716\u4e2d\u7684\u97f3\u7d20\uff0c\u4e5f\u8a31\u5728\u6211\u5011 39 \u500b models \u8a0a\u865f\u7684\u5167\u5bb9\u548c\u6a19\u6e96\u8a9e\u97f3\u8a0a\u865f\u7684\u5167\u5bb9\u76f8\u540c\u300d\u9019\u53e5\u8a71\u662f\u76f8\u7576\u53ef\u9760\u7684\uff0c\u4e5f\u5c31\u8868\u793a\u6211\u5011\u53ef\u4ee5\u653e\u5fc3\u5730\u91dd\u5c0d\u9019\u53e5\u8a9e\u97f3\u8a0a\u865f\u9032 \u7684\u767c\u97f3\u76f8\u4f3c\uff0c\u800c\u6709\u4e9b\u97f3\u7d20\u7684\u767c\u97f3\u5dee\u7570\u5247\u76f8\u7576\u5927[16]\uff0c\u56e0\u6b64\u6211\u5011\u5c0d\u65bc\u4e0a\u65b9\u5716\u4e2d\u7684\u97f3\u7d20\uff0c\u53ef\u89e3\u91cb\u6210\u5176\u7b2c\u4e00\u540d\u548c\u7b2c\u4e8c \u5c0d\u65bc\u5728\u5be6\u9a57\u4e2d\u6c42\u51fa\u7684\u9580\u6abb\u503c(threshold)\u800c\u8a00\uff0c\u5982\u679c\u8a9e\u97f3\u8a0a\u865f\u5f97\u5230\u7684\u53ef\u4fe1\u5ea6\u503c\u9ad8\u65bc\u9580\u6abb\u503c\uff0c\u5247\u6211\u5011\u7a31\u300c\u6b64\u53e5\u8a9e\u97f3 \u7684\u65b9\u5f0f\uff0c\u5c07\u8a9e\u97f3\u76e1\u53ef\u80fd\u5730\u4f9d\u5e8f\u5207\u5272\u51fa\u6bcf\u4e00\u500b\u97f3\u7d20\u3002\u5728\u9019\u7a2e\u60c5\u6cc1\u4e0b\uff0c\u8a55\u5206\u8a9e\u97f3\u5207\u5272\u5f8c\uff0c\u5982\u679c\u539f\u4f86\u7684\u5167\u5bb9\u548c\u6a19\u6e96\u8a9e uw\uff0c\u5c31\u7121\u6cd5\u518d\u7e7c\u7e8c\u3002\u5716\u8868 3\u4e4b\u97f3\u7d20\u7b26\u865f\u662f\u63a1\u7528 CMU Phone Set \u8868\u793a\u6cd5[21]\u3002 sil sh iy hh ae d y ao r d aa r k s uw t ah n g r iy s iy w aa sh w ao t er ao l y ih r sil \u22120.2 \u22120.1 0 0.1 0.2 sil sh iy hh ae d y ao r d aa r k s uw \u22120.2 \u22120.1 0 0.1 0.2 \u5716\u8868 3 \u8aaa\u8a71\u9a57\u8b49\u7684\u8a9e\u97f3\u8a0a\u865f\u5207\u5272\u6bd4\u8f03\u5716 3.1.2 \u97f3\u7d20\u6392\u540d \u5207\u5272\u8a9e\u97f3\u8a0a\u865f\u5f97\u5230\u97f3\u7d20\u6642\u9593\u5340\u6bb5\u5f8c\uff0c\u9996\u5148\u4ee5\u6bcf\u500b\u97f3\u7d20\u5c0d 39 \u500b phone models \u8a08\u7b97\u5c0d\u6578\u6a5f\u7387[21]\uff0c\u4e26\u4ee5\u6392\u540d\u7684\u9806\u5e8f \u5f97\u5230\u76f8\u5c0d\u61c9\u7684\u53ef\u4fe1\u5ea6\u503c\u3002\u6a5f\u7387\u6392\u540d\u7684\u793a\u610f\u5716\u5982\u5716\u8868 4\uff1a \u5716\u8868 4 \u97f3\u7d20\u6a5f\u7387\u6392\u540d \u4e0a\u4e0b\u5169\u500b\u6a5f\u7387\u5206\u4f48\u8868\u793a\u4e0d\u540c\u7684\u97f3\u7d20\u7d93\u7531\u8fa8\u8b58\u7a0b\u5f0f\u6c42\u5f97 39 \u500b\u5c0d\u6578\u6a5f\u7387\u7684\u7d50\u679c\uff0c\u7531\u5716\u8868 4\u53ef\u4ee5\u770b\u51fa\uff0c\u5c0d\u65bc\u4e0d \u60c5\u6cc1\uff0c\u4e5f\u5c31\u662f\u5982\u4f55\u5c07\u97f3\u7d20\u7684\u6392\u540d\u6b63\u898f\u5316\uff0c\u5f97\u5230\u4e00\u500b\u5408\u7406\u7684\u6578\u503c\u3002 \u5728 Sukkar \u548c ( ) \uf8f7 \uf8f7 \uf8f8 \uf8f6 \uf8ec \uf8ec \uf8ed \uf8eb \u22c5 \u2212 \u22c5 + = 1 log log 1 exp 1 2 Rank Rank pho pho P P Rank value pho \u03b1 ( ) x exp \u8868\u793a x e \uff0c\u5373\u81ea\u7136\u5c0d\u6578\u7684 e \u7684 x \u6b21\u65b9\u3002 pho Rank \u548c pho Rank P log \u5206\u5225\u8868\u793a\u8a72\u97f3\u7d20\u5728 39 \u500b models \u4e2d\u7684\u6392\u540d\u53ca\u5c0d \u6578\u6a5f\u7387\u503c\uff0c1 \u8868\u793a\u7b2c\u4e00\u540d\uff0c\u03b1 \u70ba\u6211\u5011\u8abf\u6574\u7684\u53c3\u6578\u503c\u3002\u7531\u6b64\u516c\u5f0f\u53ef\u5f97\u77e5\uff0c\u7576\u67d0\u97f3\u7d20\u76f8\u5c0d\u65bc 39 \u500b models \u7684\u6392\u540d\u70ba \u7b2c\u4e00\u540d\u6642\uff0c\u8a72\u97f3\u7d20\u7684\u53ef\u4fe1\u5ea6\u503c\u70ba 1\u3002 \u5716\u8868 5 \u8868\u793a\u5c0d\u65bc\u300cSH\u300d\u9019\u500b\u97f3\u7d20\u4e4b\u8a9e\u97f3\u5340\u6bb5\u85c9\u7531\u4e0a\u8ff0\u7684\u516c\u5f0f\u53ef\u5c07\u5176\u5c0d\u61c9\u65bc 39 \u500b models \u6240\u7522\u751f\u7684\u5c0d\u6578\u6a5f \u7387\u53ca\u540d\u6b21\u63db\u7b97\u6210\u53ef\u4fe1\u5ea6\u503c\u3002\u5f9e\u5716\u4e2d\u53ef\u4ee5\u770b\u51fa\uff0c\u7576\u540d\u6b21\u5728\u7b2c 10 \u540d\u5de6\u53f3\u6642\uff0c\u53ef\u4fe1\u5ea6\u503c\u5df2\u7d93\u964d\u81f3 0.2 \u4e86\u3002 \u5716\u8868 5 \u97f3\u7d20 SH \u7684\u6392\u540d\u8207\u53ef\u4fe1\u5ea6\u503c\u7684\u95dc\u4fc2 \u53e6\u5916\u7531\u65bc\u97f3\u7d20\u9593\u767c\u97f3\u7684\u5dee\u7570\u6027\uff0c\u56e0\u6b64\u6211\u5011\u5728\u8a55\u65b7\u53ef\u4fe1\u5ea6\u503c\u6642\uff0c\u4e0d\u80fd\u55ae\u7d14\u5730\u4ee5\u6392\u540d\u4f86\u505a\u6bd4\u8f03\u3002\u8209\u4f8b\u4f86\u8aaa\uff0c\u97f3 \u7d20\u300cOW\u300d\u3014o\u3015\u548c\u300cS\u300d\u3014s\u3015\u6bd4\u5c0d\u5b8c 39 \u500b models \u5f8c\u540c\u6a23\u90fd\u5f97\u5230\u7b2c\u4e8c\u540d\u7684\u7d50\u679c\uff0c\u4f46\u662f\u5c0d\u65bc\u300cOW\u300d\u800c\u8a00\uff0c\u5176\u7b2c \u4e00\u540d\u662f\u300cAO\u300d\u3014R\u3015\uff0c\u800c\u300cS\u300d\u97f3\u7d20\u7684\u7b2c\u4e00\u540d\u662f\u300cT\u300d\u3014t\u3015\uff0c\u5247\u6211\u5011\u53ef\u4ee5\u5f88\u660e\u986f\u5730\u770b\u51fa\u300cOW\u300d\u548c\u7b2c\u4e00\u540d\u7684\u5c0d\u6578 \u6a5f\u7387\u5dee\u8ddd\u8f03\u5c0f\uff0c\u4e5f\u56e0\u6b64\u53ef\u4fe1\u5ea6\u503c\u61c9\u8a72\u8981\u6bd4\u8f03\u9ad8\u624d\u5408\u7406\u3002\u56e0\u6b64\u5728\u4e0a\u8ff0\u516c\u5f0f\u4e2d\uff0c\u6211\u5011\u5c07\u6392\u540d\u7684\u5dee\u7570\u518d\u4e58\u4e0a\u5c0d\u6578\u6a5f\u7387 \u7684\u6bd4\u4f8b\u5dee\u7570\uff0c\u5982\u6b64\u4e00\u4f86\u5c31\u6703\u4f7f\u5f97\u6bcf\u500b\u97f3\u7d20\u7684\u53ef\u4fe1\u5ea6\u503c\u53d7\u5230\u6392\u540d\u53ca\u5c0d\u6578\u6a5f\u7387\u7684\u5f71\u97ff\u3002\u6700\u5f8c\u7d93\u7531\u8a08\u7b97\u5f97\u5230\u7684\u53ef\u4fe1\u5ea6 \u503c\u4ecb\u65bc 0 \u548c 1 \u4e4b\u9593\u3002 \u7576\u8a08\u7b97\u51fa\u53e5\u5b50\u6240\u6709\u6210\u529f\u5207\u5272\u7684\u97f3\u7d20\u53ef\u4fe1\u5ea6\u503c\u4e4b\u5f8c\uff0c\u5229\u7528\u6bcf\u500b\u97f3\u7d20\u7684\u6642\u9593\u9577\u5ea6\u5360\u53e5\u5b50\u6642\u9593\u9577\u5ea6\u7684\u767e\u5206\u6bd4\u4f5c\u70ba \u6b0a\u91cd\uff0c\u5373\u53ef\u63a8\u5c0e\u5f97\u51fa\u4e00\u53e5\u8a9e\u97f3\u8a0a\u865f\u7684\u53ef\u4fe1\u5ea6\u503c\u3002\u4ee5\u4e0b\u662f\u8a2d\u5b9a\u7684\u516c\u5f0f\uff1a ( ) ( ) \u2211 = \u22c5 \u22c5 = N n pho n sen n value sentence len pho len value 1 100 , N \u70ba\u4e00\u55ae\u5b57\u4e2d\u8a55\u5206\u97f3\u7d20\u7684\u6578\u91cf\uff0c ( ) x len \u8868\u793a x \u7684\u6642\u9593\u9577\u5ea6\u3002 \u81f3\u65bc\u6709\u4e9b\u55ae\u5b57\u53ef\u80fd\u5176\u4e2d\u7684\u4e00\u4e9b\u97f3\u7d20\u6c92\u6709\u8fa6\u6cd5\u7d93\u7531\u8a9e\u97f3\u8a0a\u865f\u5207\u5272\u7522\u751f\uff0c\u5c0d\u65bc\u9019\u4e9b\u97f3\u7d20\uff0c\u6211\u5011\u5c31\u76f4\u63a5\u5c07\u5176 pho value \u8a2d\u70ba 0\u3002\u6700\u5f8c\u4e58\u4e0a\u5e38\u6578 100 \u4ee3\u8868\u6211\u5011\u5c07\u8aaa\u8a71\u9a57\u8b49\u7cfb\u7d71\u7684\u7d50\u679c\u5b9a\u7fa9\u5728 0 \u81f3 100 \u4e4b\u9593\u3002 3.1.4 \u8aaa\u8a71\u9a57\u8b49\u5be6\u9a57\u7d50\u679c \u5be6\u9a57\u7528\u7684\u8a9e\u6599\u5176\u97f3\u8a0a\u683c\u5f0f\u7686\u70ba PCM\uff0c\u97f3\u8a0a\u53d6\u6a23\u983b\u7387\u70ba 16 kHz\uff0c\u4f4d\u5143\u89e3\u6790\u5ea6\u70ba 16 bits\uff0c\u6240\u6709\u7684\u5be6\u9a57\u8a9e\u6599\u7686 \u70ba\u55ae\u8072\u9053\u3002\u63a5\u8457\u5c07\u4e0a\u8ff0\u5169\u90e8\u4efd\u5404 168 \u53e5\u7684\u5be6\u9a57\u8a9e\u6599\u7d93\u7531\u8aaa\u8a71\u9a57\u8b49\u7cfb\u7d71\u5f97\u5230\u5c0d\u61c9\u7684\u53ef\u4fe1\u5ea6\u503c\uff0c\u800c\u5f8c\u518d\u7d71\u8a08\u3001\u5206\u6790 \u9019\u4e9b\u53ef\u4fe1\u5ea6\u503c\u5373\u6c42\u5f97\u9a57\u8b49\u7cfb\u7d71\u7684\u9580\u6abb\u503c\u3002\u5716\u8868 6 \u70ba\u6c42\u53d6\u9580\u6abb\u503c\u7684\u5be6\u9a57\u7d50\u679c\u5206\u4f48\u5716\uff0c\u6a6b\u8ef8\u70ba\u53ef\u4fe1\u5ea6\u503c\u7684\u7bc4\u570d\uff0c \u7e31\u8ef8\u70ba\u53ef\u4fe1\u5ea6\u503c\u8655\u65bc\u8a72\u7bc4\u570d\u5167\u7684\u8a9e\u97f3\u8a0a\u865f\u500b\u6578\u3002 \u5716\u8868 6 \u8aaa\u8a71\u9a57\u8b49\u6c42\u53d6\u9580\u6abb\u503c\u5be6\u9a57\u7d50\u679c\u5206\u4f48\u60c5\u6cc1 \u6211\u5011\u4ee5\u300c\u578b\u5225 I 3.2 \u8a9e\u97f3\u8a0a\u865f\u5207\u5272 \u300c\u8a9e\u97f3\u8a0a\u865f\u5207\u5272\u300d\u6a21\u7d44\u7684\u529f\u80fd\u4e43\u662f\u5c07\u6a19\u6e96\u8a9e\u6599\u53ca\u8a55\u5206\u8a9e\u6599\u5207\u5272\u51fa\u97f3\u7d20\u767c\u97f3\u7684\u5340\u6bb5\u3002\u5176\u4f5c\u6cd5\u662f\u4ee5\u9810\u5148\u8a13\u7df4\u597d\u7684\u82f1 \u6587\u767c\u97f3\u8072\u5b78\u6a21\u578b\uff0c\u5207\u5272\u51fa\u8a9e\u6599\u4e2d\u4e4b\u6b63\u78ba\u7684\u97f3\u7d20\u767c\u97f3\u5340\u6bb5\u3002\u4ee5\u4e0b\u7ae0\u7bc0\u5c07\u5206\u6210\u300c\u8072\u5b78\u6a21\u578b\u7684\u8a13\u7df4\u300d\u548c\u300c\u5229\u7528\u8a9e\u97f3\u8fa8 \u8b58\u4f86\u9032\u884c\u8a9e\u97f3\u8a0a\u865f\u5207\u5272\u300d\u9019\u5169\u90e8\u4efd\u4f86\u4ecb\u7d39\u3002 3.2.1 3.2.2 \u8072\u5b78\u6a21\u578b\u8a2d\u8a08 \u82f1\u6587\u4e2d\u6bcf\u4e00\u500b\u97f3\u7bc0\u53ef\u80fd\u7531\u4e00\u500b\u6216\u6578\u500b\u97f3\u6a19\u6240\u7d44\u6210\uff0c\u800c\u6bcf\u4e00\u500b\u97f3\u6a19\u90fd\u6703\u5c0d\u61c9\u5230\u4e00\u500b\u97f3\u7d20\uff0c\u800c\u8072\u8abf\u3001\u91cd\u97f3\u548c\u7834\u97f3 \u97f3\u6a19 \u6a21\u578b \u97f3\u6a19 \u6a21\u578b \u97f3\u6a19 \u6a21\u578b \u97f3\u6a19 \u6a21\u578b \u97f3\u6a19 (multiple pronunciation)\u6a21\u578b AA</td></tr></table>",
                "html": null,
                "type_str": "table",
                "text": "Computer-Assisted Language Learning)\u5df2\u53d7\u5230\u76f8\u7576\u91cd\u8996\uff0c\u5404\u65b9\u4e5f\u7d1b\u7d1b\u6295\u5165\u76f8\u95dc\u7684\u7814\u7a76[10][11][18][15][20]\u3002 \u96fb\u8166\u8f14\u52a9\u767c\u97f3\u8a13\u7df4(CAPT, Computer-Assisted Pronunciation Training)\u53ef\u8996\u70ba\u662f\u8a9e\u97f3\u8fa8\u8b58\u548c\u5716\u5f62\u6bd4\u5c0d(Pattern Matching)\u5169\u9805\u6280\u8853\u7684\u7d50\u5408\u3002\u672c\u8ad6\u6587\u7814\u7a76\u4e3b\u984c\uff0c\u5305\u542b\u300c\u8aaa\u8a71\u9a57\u8b49\u300d\u3001\u300c\u8a9e\u97f3\u8a0a\u865f\u5207\u5272\u300d\u4ee5\u53ca\u300c\u82f1\u6587\u8a9e\u97f3\u8a55\u5206\u300d\u4e09 \u500b\u90e8\u4efd\uff0c\u5e0c\u671b\u878d\u5408\u76ee\u524d\u8a9e\u97f3\u8fa8\u8b58\u548c\u5716\u5f62\u6bd4\u5c0d\u7684\u6280\u8853\uff0c\u5c0d\u4f7f\u7528\u8005\u9032\u884c\u516c\u6b63\u7684\u8a9e\u97f3\u8a55\u5206\u3002 \u5728\u8a9e\u97f3\u8a55\u5206\u7cfb\u7d71\u4e2d\uff0c\u5982\u679c\u80fd\u5148\u6ffe\u9664\u5167\u5bb9\u548c\u6a19\u6e96\u8a9e\u97f3\u5b8c\u5168\u4e0d\u540c\u7684\u8a55\u5206\u8a9e\u97f3\uff0c\u53ef\u4ee5\u4f7f\u6574\u500b\u8a9e\u97f3\u8a55\u5206\u7cfb\u7d71\u66f4\u5177 \u516c\u4fe1\u529b\u3002\u672c\u8ad6\u6587\u904b\u7528\u4e86\u53ef\u4fe1\u5ea6\u8a55\u4f30\u7684\u6280\u8853\u4f86\u9054\u6210\u8aaa\u8a71\u9a57\u8b49(Utterance Verification)\u3002\u78ba\u4fdd\u4e86\u8a55\u5206\u8a9e\u97f3\u5167\u5bb9\u7684\u6b63\u78ba \u6027\u5f8c\uff0c\u5c0d\u65bc\u8a55\u5206\u8a9e\u97f3\u6211\u5011\u4f7f\u7528 HMM(Hidden Markov Model)\u5207\u5272\u51fa\u6bcf\u500b\u97f3\u7d20(phoneme)\u7684\u6642\u9593\u5340\u6bb5\uff0c\u4f7f\u7528\u9ad8\u8fa8\u8b58 \u7387\u7684 HMM \u8072\u5b78\u6a21\u578b\u53ef\u78ba\u4fdd\u5207\u5272\u51fa\u4f86\u7684\u97f3\u7d20\u5340\u6bb5\u6709\u4e00\u5b9a\u7684\u53ef\u4fe1\u5ea6\u53ca\u6b63\u78ba\u7387\u3002\u5728\u82f1\u6587\u8a9e\u97f3\u8a55\u5206\u90e8\u4efd\uff0c\u6211\u5011\u5229\u7528\u6a19 \u6e96\u8a9e\u97f3\u8cc7\u6599\u4f86\u9032\u884c\u4e00\u7a2e\u8f03\u70ba\u4e3b\u89c0\u7684\u8a55\u5206\u65b9\u5f0f\uff0c\u4e3b\u8981\u4f7f\u7528\u5716\u6a23\u6bd4\u5c0d(Pattern Matching)\u7684\u65b9\u6cd5\uff0c\u6839\u64da\u56db\u500b\u8a55\u5206\u53c3\u6578\uff1a \u97f3\u91cf\u5f37\u5ea6\u66f2\u7dda(Magnitude) \u3001\u57fa\u983b\u8ecc\u8de1\u66f2\u7dda(Pitch Contour) \u3001\u767c\u8072\u6025\u7de9\u8b8a\u5316(Rhythm)\u4ee5\u53ca HMM \u5c0d\u6578\u6a5f\u7387\u5dee\u7570(HMM Log-Likelihood)\uff0c\u5c07\u8a55\u5206\u8a9e\u97f3\u548c\u6a19\u6e96\u8a9e\u97f3\u7684\u8cc7\u6599\u9010\u97f3\u7d20\u5730\u4f86\u505a\u6bd4\u8f03\uff0c\u4ee5\u671f\u627e\u51fa\u8a55\u5206\u8a9e\u97f3\u548c\u6a19\u6e96\u8a9e\u97f3\u7684\u5dee\u7570\u7a0b\u5ea6\u3002 2 \u76f8\u95dc\u7814\u7a76 1997 \u5e74\u6642\uff0cC. Cucchiarini\u3001H. Strik \u53ca L. Boves \u4ee5\u8377\u862d\u8a9e\u70ba\u4e3b\uff0c\u5b9a\u7fa9\u4e86 Total Duration of Speech no/plus Pause\u3001 Mean Segment Duration \u3001 Rate of Speech \u4ee5 \u53ca Global Log-Likelihood \uff0c \u7d93 \u7531 \u985e \u4f3c \u7684 \u5be6 \u9a57 \u5f8c \u5f97 \u51fa Global Log-Likelihood \u5c0d\u65bc\u4eba\u985e\u4e3b\u89c0\u8a55\u5206\u5360\u8f03\u91cd\u7684\u6bd4\u91cd[19]\u30021999 \u5e74 L. Neumeyer\u3001H. Franco\u3001V. Digalakis \u548c M. Weintraub \u4ee5\u6cd5\u8a9e\u8a9e\u6599\u5eab\u9032\u884c\u5be6\u9a57\uff0c\u63a1\u7528 HMM Log-Likelihood\u3001Normalized Acoustic\u3001Segment classification\u3001 Segment Duration\u3001Timing \u7576\u4f5c\u5176\u5be6\u9a57\u7684\u8a55\u5206\u53c3\u6578\uff0c\u7d93\u7531\u5be6\u9a57\u5f8c\u5f97\u51fa\u4e86 Normalized Acoustic \u5728\u8a55\u5206\u7cfb\u7d71\u548c\u8a9e\u8a00 \u5c08\u5bb6\u7d66\u4e88\u7684\u5206\u6578\u4e2d\uff0c\u5176\u76f8\u95dc\u6027\u9ad8\u65bc Lee \u65bc 1996 \u5e74\u767c\u8868\u7684\u8ad6\u6587[17]\u4e2d\u63d0\u5230\uff0c\u97f3\u7d20\u7684\u5c0d\u6578\u6a5f\u7387\u4ee5\u53ca\u5c0d\u6240\u6709\u97f3\u7d20\u7684\u5c0d\u6578\u6a5f\u7387\u6392\u540d\uff0c\u548c \u9a57\u8b49\u7cfb\u7d71\u7684\u53ef\u4fe1\u5ea6\u503c\u662f\u6709\u5f88\u5927\u5f71\u97ff\u7684\u3002\u57fa\u65bc\u4ee5\u4e0a\u7684\u524d\u63d0\uff0c\u6211\u5011\u5c07 Sukkar \u548c Lee \u6240\u63d0\u51fa\u6c42\u53d6\u53ef\u4fe1\u5ea6\u503c\u7684\u5f0f\u5b50\u6539\u5beb \u4e26\u4ee5\u4e0b\u5217\u516c\u5f0f\u8868\u793a\uff1a \u932f\u8aa4\u7387(Type I error, False Reject)\u52a0\u4e0a\u578b\u5225 II \u932f\u8aa4\u7387(Type II error, False Accept)\u70ba\u6700\u5c0f\u300d\u4f5c \u70ba\u5c0b\u627e\u9580\u6abb\u503c\u7684\u524d\u63d0\u3002\u6839\u64da\u5be6\u9a57\u7d50\u679c\uff0c\u6211\u5011\u767c\u73fe Correct \u4e2d\u7684\u8a9e\u6599\u5176\u6700\u5c0f\u53ef\u4fe1\u5ea6\u503c\u70ba 63.21\uff0c\u800c\u5728 Incorrect \u53ef\u4fe1 \u5ea6\u503c\u5927\u65bc 60 \u7684\u8a9e\u6599\u4e2d\u6700\u63a5\u8fd1 63.21 \u7684\u53ef\u4fe1\u5ea6\u503c\u70ba 61.59\uff0c\u56e0\u6b64\u6211\u5011\u5c07\u8aaa\u8a71\u9a57\u8b49\u7cfb\u7d71\u7684\u9580\u6abb\u503c\u8a2d\u5b9a\u6210 62.40(\u5373\u5169 \u8005\u7684\u5e73\u5747)\uff0c\u5982\u6b64\u53ef\u9054\u5230\u578b\u5225 I \u932f\u8aa4\u7387\u70ba 0%\uff0c\u578b\u5225 II \u932f\u8aa4\u7387\u70ba\u70ba 1.19%\u3002 \u7d93\u7531\u4e0a\u8ff0\u5be6\u9a57\u8a08\u7b97\u6c42\u51fa\u9580\u6abb\u503c\u5f8c\uff0c\u6211\u5011\u53e6\u5916\u6e96\u5099\u4e00\u7d44\u5167\u542b Correct \u53ca Incorrect \u5404\u70ba 168 \u53e5\u7684\u6e2c\u8a66\u8a9e\u6599\uff0c\u5176 \u4e2d Correct \u8a9e\u6599\u7684\u8a9e\u6599\u9577\u5ea6\u7e3d\u548c\u7d04\u70ba 7 \u5206 27 \u79d2\uff0cIncorrect \u8a9e\u6599\u7684\u8a9e\u6599\u9577\u5ea6\u7e3d\u548c\u7d04\u70ba 8 \u5206 57 \u79d2\u3002\u5c07\u9019\u4e9b\u8a9e\u6599\u4ee5\u9580 \u6abb\u503c\u70ba 62.40 \u7684\u5be6\u9a57\u7d50\u679c\uff0c\u5176\u578b\u5225 I \u932f\u8aa4\u7387\u70ba 7.14%\uff0c\u578b\u5225 II \u932f\u8aa4\u7387\u70ba\u70ba 0.60%\u3002 \u8072\u5b78\u6a21\u578b HMM \u7684\u8a9e\u6599 \u5be6\u4f5c\u8a9e\u97f3\u8a0a\u865f\u5207\u5272\u4e4b\u524d\uff0c\u6211\u5011\u5fc5\u9808\u5148\u7522\u751f\u8072\u5b78\u6a21\u578b\uff0c\u624d\u80fd\u91dd\u5c0d\u5404\u7a2e\u4e0d\u540c\u7684\u8a9e\u97f3\u9032\u884c\u5207\u5272\u52d5\u4f5c\u3002\u672c\u8ad6\u6587\u4e2d\u6211\u5011\u8a2d \u8a08\u4e86\u5169\u7a2e\u4e0d\u540c\u7684\u8072\u5b78\u6a21\u578b\uff1a\u4e00\u500b\u662f\u81fa\u7063\u4eba\u53e3\u97f3\u7684\u8072\u5b78\u6a21\u578b\uff0c\u4e00\u500b\u662f\u5916\u570b\u4eba\u6a19\u6e96\u8a9e\u97f3\u7684\u8072\u5b78\u6a21\u578b\u3002 \u9996\u5148\u91dd\u5c0d\u6bcd\u8a9e\u70ba\u82f1\u6587\u7684\u8072\u5b78\u6a21\u578b\uff0c\u6211\u5011\u4f7f\u7528 TIMIT \u8a9e\u6599\u4f86\u52a0\u4ee5\u8a13\u7df4\u3002\u8a9e\u6599\u5167\u5bb9\u70ba 2,342 \u53e5\u5e73\u8861\u8a9e\u6599\uff0c\u7531 438 \u4f4d\u7537\u6027\u3001192 \u4f4d\u5973\u6027\uff0c\u5171 630 \u4eba\u9304\u88fd\uff0c\u6bcf\u4eba\u5206\u914d\u9304\u88fd 10 \u53e5\uff0c\u6545\u5171\u6709 6,300 \u53e5\u8a9e\u97f3\u3002\u4f9d TIMIT \u7684\u5efa\u8b70\u53d6\u5176\u4e2d 4,620 \u53e5\u3001\u8a9e\u6599\u9577\u5ea6\u7e3d\u548c\u7d04\u70ba 3 \u5c0f\u6642 49 \u5206 10 \u79d2\u7684\u8a9e\u97f3\u8a0a\u865f\u4f5c\u70ba\u6bcd\u8a9e\u70ba\u82f1\u6587\u7684\u8072\u5b78\u6a21\u578b\u8a13\u7df4\uff0c\u53e6\u5916 1,680 \u53e5\u3001\u8a9e \u6599\u9577\u5ea6\u7e3d\u548c\u7d04\u70ba 1 \u5c0f\u6642 23 \u5206 51 \u79d2\u7684\u8a9e\u97f3\uff0c\u5247\u4f5c\u70ba\u5916\u5728\u6e2c\u8a66\u6a94(Outside Test)\u3002 \u53e6\u4e00\u65b9\u9762\u91dd\u5c0d\u6bcd\u8a9e\u70ba\u570b\u8a9e\u7684\u8072\u5b78\u6a21\u578b\uff0c\u6211\u5011\u8acb 33 \u4f4d\u5b78\u751f\uff0c\u5176\u4e2d\u5305\u542b\u4e86 23 \u4f4d\u7537\u6027\u300110 \u4f4d\u5973\u6027\uff0c\u4f9d TIMIT \u7684\u8cc7\u6599\u9304\u88fd 7,026 \u53e5\u5e73\u8861\u8a9e\u6599\uff0c\u6211\u5011\u53d6\u5176\u4e2d\u7684 4,684 \u53e5\u3001\u8a9e\u6599\u9577\u5ea6\u7e3d\u548c\u7d04\u70ba 4 \u5c0f\u6642 11 \u5206 3 \u79d2\u7684\u8a9e\u97f3\u4f5c\u70ba\u6bcd\u8a9e\u70ba \u4e2d\u6587\u7684\u8072\u5b78\u6a21\u578b\u8a13\u7df4\uff0c\u800c\u53e6\u5916\u7684 2,342 \u53e5\u3001\u8a9e\u6599\u9577\u5ea6\u7e3d\u548c\u7d04\u70ba 1 \u5c0f\u6642 57 \u5206 43 \u79d2\u7684\u8a9e\u97f3\u4f5c\u70ba\u5916\u5728\u6e2c\u8a66\u6a94\u3002 \u4e0a\u8ff0\u8a9e\u6599\u7684\u97f3\u8a0a\u683c\u5f0f\u7686\u70ba PCM\uff0c\u53d6\u6a23\u983b\u7387\u70ba 16 kHz\uff0c\u4f4d\u5143\u89e3\u6790\u5ea6\u70ba 16 bits\u3002 \u7684\u554f\u984c\uff0c\u5728\u76ee\u524d\u7684\u8072\u5b78\u6a21\u578b\u8a2d\u8a08\u4e2d\u5247\u66ab\u6642\u5ffd\u7565\u3002TIMIT \u7684\u5b57\u5178\u6709 62 \u500b\u97f3\u7d20\uff0c\u7531\u65bc\u83ef\u4eba \u5c0d\u65bc\u4e00\u4e9b\u97f3\u7d20\u4e0d\u50cf\u5916\u570b\u4eba\u5ff5\u5f97\u90a3\u9ebc\u6e96\u78ba\uff0c\u518d\u52a0\u4e0a\u8a13\u7df4\u8a9e\u6599\u4e0d\u8db3\u4e0b\uff0c\u5982\u679c\u6211\u5011\u6e1b\u5c11\u8a13\u7df4 model \u7684\u500b\u6578\uff0c\u5247\u53ef\u4f7f\u6bcf \u500b model \u7684\u8a13\u7df4\u8a9e\u6599\u53d6\u6a23\u6578\u76ee\u589e\u591a\u3002\u9451\u65bc\u4e0a\u8ff0\u5169\u500b\u539f\u56e0\uff0c\u6211\u5011\u5c07\u539f\u5148 TIMIT \u8a2d\u8a08\u7684 62 \u500b\u97f3\u7d20\u522a\u6e1b\u6210 40 \u500b\u97f3\u7d20 (\u542b\u975c\u97f3 SIL \u97f3\u7d20)\u3002\u5728\u672c\u7ae0\u4e2d\u6211\u5011\u4f7f\u7528\u7684\u8072\u5b78\u6a21\u578b\u548c\u97f3\u7d20\u662f\u4e00\u5c0d\u4e00\u5c0d\u61c9\u7684\u3002\u8209\u4f8b\u4f86\u8aaa\uff0c\"school\"\u9019\u500b\u55ae\u5b57\uff0c\u5176 KK \u97f3\u6a19\u70ba\u3014skul\u3015\uff0c\u4ee5\u6211\u5011\u8a2d\u8a08\u7684\u8072\u5b78\u6a21\u578b\u4f86\u8aaa\uff0c\u5c31\u662f\u300cS\u300d\uff0b\u300cK\u300d\uff0b\u300cUW\u300d\uff0b\u300cL\u300d\u3002\u8868\u683c 1 \u662f\u6211\u5011\u6240 \u8a2d\u8a08\u7684 40 \u500b\u8072\u5b78\u6a21\u578b\u8207 KK \u97f3\u6a19\u5c0d\u7167\u8868\uff1a \u8868\u683c 1 40 \u500b\u8072\u5b78\u6a21\u578b\u8207 KK \u97f3\u6a19\u5c0d\u7167\u8868",
                "num": null
            },
            "TABREF1": {
                "content": "<table><tr><td>Hidden Markov \u8a9e\u97f3\u8a0a\u865f\u5207\u5272\u7684\u4e3b\u8981\u76ee\u6a19\u5373\u662f\u5e0c\u671b\u80fd\u5920\u5c07\u9023\u7e8c\u7684\u82f1\u6587\u8a9e\u97f3\u53e5\u5b50\uff0c\u5176\u4e2d\u5305\u542b\u4e86\u6a19\u6e96\u8a9e\u97f3\u548c\u8a55\u5206\u7684\u8a9e\u97f3\uff0c\u5207\u5272\u6210\u7368 Model Toolkit)\u9032\u884c\u8a13\u7df4\u3002 3.2.4 \u8a13\u7df4\u7d50\u679c \u7acb\u7684\u97f3\u7d20\uff0c\u5982\u6b64\u4e00\u4f86\u6211\u5011\u624d\u53ef\u4ee5\u91dd\u5c0d\u6bcf\u4e00\u6bb5\u53e5\u5b50\u4e2d\u7684\u97f3\u7d20\u548c\u6a19\u6e96\u8a9e\u97f3\u4e2d\u7684\u6bcf\u4e00\u500b\u97f3\u7d20\u505a\u6bd4\u8f03\u3002\u5728\u6b64\u6211\u5011\u4f7f\u7528\u5f37 \u8feb\u5c0d\u61c9(Forced Alignment)[6]\u7684\u65b9\u5f0f\u5c07\u8a9e\u97f3\u8a0a\u865f\u5207\u5272\u6210\u5404\u500b\u97f3\u7d20\u7684\u6642\u9593\u5340\u6bb5\uff0c\u4ee5\u5229\u8a55\u5206\u6a5f\u5236\u7684\u904b\u4f5c\u3002\u5728\u524d\u8655\u7406\u7684 \u904e\u7a0b\u4e2d\uff0c\u6211\u5011\u5229\u7528\u5167\u542b 127,102 \u500b\u82f1\u6587\u55ae\u5b57\u7684 CMU \u5b57\u5178(Dictionary from Carnegie Mellon University)\u5c0d\u5404\u55ae\u5b57\u6a19 \u97f3\u4e26\u5efa\u7acb\u5404\u81ea\u7368\u7acb\u7684\u8fa8\u8b58\u7db2\u8def[21]\u3002\u5982\u4e0b\u5716\uff1a \u5716\u8868 7 \u8a9e\u97f3\u8a0a\u865f\u5207\u5272\u524d\u8655\u7406\u6d41\u7a0b\u793a\u610f\u5716 \u5b8c\u6210\u524d\u8655\u7406\u52d5\u4f5c\u5f8c\uff0c\u6211\u5011\u53ef\u7e7c\u7e8c\u9032\u884c\u8a9e\u97f3\u8a0a\u865f\u5207\u5272\u7684\u6d41\u7a0b\uff0c\u9996\u5148\u5c07\u4e00\u8a9e\u97f3\u8a0a\u865f\u7d93\u904e\u7aef\u9ede\u5075\u6e2c\u5f8c\u518d\u7d93\u7531\u7279 \u5fb5\u64f7\u53d6\uff0c\u53d6\u51fa\u8a9e\u97f3\u4e2d\u7684\u7279\u5fb5\uff0c\u7136\u5f8c\u5c07\u9019\u4e9b\u7279\u5fb5\u53c3\u6578\u900f\u904e\u8072\u5b78\u6a21\u578b(\u96b1\u85cf\u5f0f\u99ac\u53ef\u592b\u6a21\u578b)\u53ca\u8a9e\u8a00\u6a21\u578b(\u8fa8\u8b58\u7db2\u8def)\uff0c\u5229 \u7528\u7dad\u7279\u6bd4\u6f14\u7b97\u6cd5(Viterbi algorithm)\u5373\u53ef\u627e\u51fa\u6700\u76f8\u4f3c\u7684\u97f3\u7d20\uff0c\u4e26\u5f97\u77e5\u5404\u97f3\u7d20\u7684\u6642\u9593\u5340\u6bb5\u3002 \u8868\u683c 2 \u8a9e\u97f3\u8a0a\u865f\u5207\u5272\u5be6\u9a57\u7d50\u679c \u5be6\u9a57\u65b9\u5f0f \u9805\u76ee N-Wave /N-HMM N-Wave /T-HMM T-Wave /N-HMM T-Wave /T-HMM \u5be6\u9a57\u8a9e\u6599\u97f3\u7d20\u7e3d\u6578 58,282 58,282 81,229 81,229 \u5207\u5272\u5f8c\u6b63\u78ba\u97f3\u7d20\u7e3d\u6578 58,253 57,142 77,293 80,230 \u97f3\u7d20\u6642\u9593\u6b63\u78ba\u7387 99.95% 98.04% 95.15% 98.77% \u5728\u5224\u65b7\u97f3\u7d20\u6642\u9593\u6b63\u78ba\u7387\u7684\u90e8\u4efd\uff0c\u5c0d\u65bc N-Wave \u800c\u8a00\uff0c\u7531\u65bc\u6240\u6709\u7684\u8a9e\u6599 TIMIT \u90fd\u6709\u63d0\u4f9b\u6a19\u97f3\u6a94\uff0c\u56e0\u6b64\u6211\u5011 \u53ef\u6bd4\u5c0d\u5207\u5272\u51fa\u4f86\u7684\u6642\u9593\u9ede\u548c\u6a19\u97f3\u6a94\uff0c\u82e5\u76f8\u5dee\u5728 0.1 \u79d2\u4ee5\u5167(5 \u500b\u97f3\u6846)\uff0c\u5247\u6211\u5011\u7a31\u6b64\u97f3\u7d20\u7684\u6642\u9593\u70ba\u6b63\u78ba\u3002\u800c\u5c0d\u65bc T-Wave \u800c\u8a00\uff0c\u7531\u65bc\u4e26\u6c92\u6709\u7d93\u904e\u4eba\u5de5\u6a19\u97f3\uff0c\u56e0\u6b64\u6211\u5011\u53ea\u5728\u9f90\u5927\u7684\u8a9e\u6599\u4e2d\u53d6\u6a23 10%\u9032\u884c\u4eba\u5de5\u5224\u65b7\uff0c\u53ea\u8981\u8a72\u5340\u6bb5\u4eba \u8033\u807d\u8d77\u4f86\u76f8\u5dee\u4e0d\u5927\uff0c\u5247\u6211\u5011\u7a31\u8a72\u97f3\u7d20\u7684\u6642\u9593\u70ba\u6b63\u78ba\u3002 \u7531\u8868\u683c 2 \u7684\u5be6\u9a57\u7d50\u679c\u53ef\u77e5\uff0c\u5728\u4e0d\u540c\u7684\u8072\u5b78\u6a21\u578b\u4e0b\uff0cForced Alignment \u7684\u97f3\u7d20\u6642\u9593\u5340\u6bb5\u90fd\u975e\u5e38\u6e96\u78ba\u3002\u8868\u683c 3 \u5247\u662f N-Wave\u3001T-Wave \u900f\u904e\u5927\u8a5e\u5f59\u8fa8\u8b58\u7684\u65b9\u5f0f\uff0c\u7d93\u7531 N-HMM\u3001T-HMM \u6240\u5f97\u51fa\u7684\u8fa8\u8b58\u7387\uff0c\u5176\u4e2d\u8a5e\u5f59\u5167\u5bb9\u70ba 2,342 \u53e5\u82f1\u6587\u53e5\u5b50\u3002 \u5be6\u9a57\u8a9e\u6599\u53e5\u5b50\u7e3d\u6578 1,680 1,680 2,342 2,342 \u8fa8\u8b58\u6b63\u78ba\u53e5\u5b50\u7e3d\u6578 1,650 622 1,997 1,425 \u53e5\u5b50\u8fa8\u8b58\u7387 98.21% 37.02% 85.26% 60.85% \u7531\u8868\u4e2d\u7684\u7d50\u679c\u6211\u5011\u53ef\u4ee5\u767c\u73fe\uff0c\u5c0d\u65bc\u76f8\u540c\u8a9e\u6599\uff0cN-HMM \u7684\u8fa8\u8b58\u7387\u7686\u9ad8\u65bc T-HMM\uff0c\u9019\u5c31\u8868\u793a\u7576\u6211\u5011\u4ee5 N-HMM \u70ba\u8072\u5b78\u6a21\u578b\u4f86\u5c0d\u8a9e\u97f3\u8a0a\u865f\u6c42\u53d6\u5c0d\u6578\u6a5f\u7387\u6642\uff0c\u6240\u5f97\u5230\u7684\u5c0d\u6578\u6a5f\u7387\u503c\u5176\u53ef\u4fe1\u5ea6\u6703\u9ad8\u65bc T-HMM\u3002\u6839\u64da\u6b64\u5be6\u9a57\u7d50\u679c\uff0c \u5728\u63a5\u4e0b\u4f86\u7684\u7ae0\u7bc0\u4e2d\uff0c\u6211\u5011\u5c07\u6703\u4ee5 N-HMM \u7576\u4f5c\u6211\u5011\u8a55\u5206\u6bd4\u5c0d\u7684\u8072\u5b78\u6a21\u578b\u3002 3.3 \u82f1\u6587\u8a9e\u97f3\u8a55\u5206 \u5716\u8868 8 \u70ba\u8a55\u5206\u7cfb\u7d71\u6d41\u7a0b\u5716\uff0c\u6211\u5011\u5c07\u5c31\u8a55\u5206\u53c3\u6578\u64f7\u53d6\u3001\u5716\u6a23\u6bd4\u5c0d\u65b9\u5f0f\u548c\u8a55\u5206\u6a5f\u5236\u5efa\u7acb\u5206\u5225\u4f5c\u4ecb\u7d39\u3002 3.3.1 \u8a55\u5206\u53c3\u6578\u64f7\u53d6 \u9664\u4e86\u97f3\u91cf\u5f37\u5ea6\u66f2\u7dda\u3001\u57fa\u983b\u8ecc\u8de1\u66f2\u7dda\u70ba\u8a55\u5206\u53c3\u6578\u5916[15]\uff0c\u6211\u5011\u4e5f\u63a1\u7528\u4e86 HMM \u5c0d\u6578\u6a5f\u7387\u5dee\u7570\u548c\u767c\u8072\u6025\u7de9\u8b8a\u5316\u9019\u5169 \u9805\u8a55\u5206\u53c3\u6578\u3002\u5728 Forced Alignment \u7684\u540c\u6642\uff0c\u6211\u5011\u53ef\u4ee5\u5f97\u5230\u6bcf\u500b\u97f3\u7d20\u5c0d\u61c9\u65bc\u8072\u5b78\u6a21\u578b\u7684\u5c0d\u6578\u6a5f\u7387(HMM log-Probability)[10][11]\u548c\u5404\u97f3\u7d20\u7684\u6642\u9593\u5340\u6bb5\uff0c\u9019\u5c31\u662f\u6240\u8b02\u7684 HMM \u5c0d\u6578\u6a5f\u7387\u5dee\u7570\u548c\u767c\u8072\u6025\u7de9\u8b8a\u5316\u9019\u5169\u9805\u8a55\u5206\u53c3 \u6578\u3002 3.3.2 \u5716\u6a23\u6bd4\u5c0d\u65b9\u6cd5 \u5728\u524d\u4e09\u500b\u8a55\u5206\u53c3\u6578\u4e2d\uff0c\u6211\u5011\u4f7f\u7528\u4e0d\u540c\u7684\u6b63\u898f\u5316\u65b9\u6cd5\u5982\u5167\u63d2\u6cd5\u3001\u7dda\u6027\u5e73\u79fb\u548c\u7dda\u6027\u7e2e\u653e[15]\uff0c\u5982\u8868\u683c 4\u3002\u800c HMM \u5c0d\u6578\u6a5f\u7387\u5dee\u7570\u5247\u63a1\u7528\u8f03\u70ba\u4e0d\u540c\u7684\u6bd4\u5c0d\u65b9\u6cd5\uff0c\u5728\u4ee5\u4e0b\u8aaa\u660e\u3002 \u8868\u683c 4 \u5404\u8a55\u5206\u53c3\u6578\u63a1\u7528\u7684\u6b63\u898f\u5316\u53ca\u8ddd\u96e2\u7b97\u6cd5 \u8a55\u5206\u53c3\u6578 \u6b63\u898f\u5316\u65b9\u6cd5 \u8ddd\u96e2\u7b97\u6cd5 \u97f3\u91cf\u5f37\u5ea6\u66f2\u7dda \u5167\u63d2\u6cd5\u3001\u7dda\u6027\u7e2e\u653e Euclidean Distance \u57fa\u983b\u8ecc\u8de1\u66f2\u7dda \u5167\u63d2\u6cd5\u3001\u7dda\u6027\u5e73\u79fb Euclidean Distance \u767c\u8072\u6025\u7de9\u8b8a\u5316 \u7121 Euclidean Distance \u5716\u8868 9 HMM \u5c0d\u6578\u6a5f\u7387\u5dee\u7570\u6bd4\u5c0d\u6d41\u7a0b\u5716 \u7531\u65bc\u6a5f\u7387\u503c\u662f\u7d55\u5c0d\u7684\uff0c\u4e0d\u5bb9\u6613\u5f9e\u6578\u503c\u76f4\u63a5\u4f5c\u6bd4\u8f03\uff0c\u56e0\u6b64\u6211\u5011\u8a2d\u8a08\u4e86\u6a5f\u7387\u500d\u6578\u4f86\u4fee\u6b63\u5c0d\u6578\u6a5f\u7387\u7684\u5dee\u7570\u503c\uff0c\u7576 \u5169\u8a9e\u97f3\u7684\u5c0d\u6578\u6a5f\u7387\u7d55\u5c0d\u503c\u7686\u5c0f\u65bc 1050 \u6642\uff0c\u6a5f\u7387\u500d\u6578\u7684\u8b8a\u5316\u8da8\u52e2\u8f03\u5c0f\u3002\u7576\u5169\u8a9e\u97f3\u7684\u5c0d\u6578\u6a5f\u7387\u7d55\u5c0d\u503c\u7686\u5927\u65bc 1050 \u6642\uff0c\u6a5f\u7387\u500d\u6578\u7684\u8b8a\u5316\u8da8\u52e2\u8f03\u5927\u3002\u95dc\u65bc\u6a5f\u7387\u500d\u6578\u6211\u5011\u5b9a\u7fa9\u4ee5\u4e0b\u7684\u516c\u5f0f\uff1a log log 1050 , 1400 log , 1 min 3 1050 0 , 350 log abs y probabilit abs y probabilit Const \uf8f4 \uf8f4 \uf8f3 \uf8f4 \uf8f4 \uf8f2 \uf8f1 &gt; \uf8f7 \uf8f7 \uf8f8 \uf8f6 \uf8ec \uf8ec \uf8ed \uf8eb \uf8fa \uf8fa \uf8f9 \uf8ef 4 3 2 1 4 3 2 1 , , , , , , , b b b b a a a a \u548c\u56db\u500b\u8a55\u5206\u53c3\u6578\u7684\u6b0a\u91cd 4 3 2 1 , , , w w w \u3002 w \uf8ef \uf8ee \u2212 + \u2264 \u2264 \uf8fa \uf8fa \uf8f9 \uf8ef \uf8ef \uf8ee \u2212 = \u8868\u683c 5 \u4eba\u5de5\u8a55\u5206\u548c\u7cfb\u7d71\u8a9e\u97f3\u8a55\u5206\u7684\u95dc\u4fc2\u5c0d\u7167\u8868 3.3.3 \u8a55\u5206\u6a5f\u5236\u5efa\u7acb \u5728\u97f3\u7d20\u5c64\u6b21\uff0c\u6211\u5011\u7531\u56db\u7a2e\u8a55\u5206\u53c3\u6578\u5f97\u5230\u4e0d\u540c\u7684\u5206\u6578\uff0c\u518d\u5f80\u4e0a\u7531\u55ae\u5b57(word)\u548c\u53e5\u5b50(sentence)\u5c64\u6b21\u4f5c\u8a55\u5206\uff0c\u5c31\u53ef\u4ee5 \u5f97\u5230\u6700\u5f8c\u8a55\u5206\u7684\u7d50\u679c\uff0c\u4ee5\u4e0b\u5247\u5206\u56db\u500b\u5c64\u6b21\u4f5c\u4ecb\u7d39\u3002 \u8a55\u5206\u53c3\u6578\u5c64\u6b21\uff1a\u5c0d\u65bc\u6bcf\u500b\u97f3\u7d20\u4e2d\u8a55\u5206\u53c3\u6578\u7684\u5206\u6578\uff0c\u6211\u5011\u8a2d\u5b9a\u4ee5\u4e0b\u7684\u516c\u5f0f[15]\uff1a ( ) b fea dist a score \u22c5 + = 1 100 \u7531\u9019\u500b\u516c\u5f0f\u6211\u5011\u5c31\u53ef\u4ee5\u5c07\u5169\u97f3\u7d20\u9593\u67d0\u500b\u7279\u5fb5\u7684\u5dee\u7570\u7a0b\u5ea6\u8f49\u6210 0 \u5230 100 \u4e4b\u9593\u7684\u5206\u6578\uff0c\u53ea\u8981\u8a2d\u5b9a\u597d\u5169\u7d44\u7684 dist \u53ca\u5c0d \u61c9\u7684 fea score \uff0c\u5373\u53ef\u5f9e\u4e2d\u6c42\u51fa a \u548c b \uff0c\u63a5\u8457\u6240\u6709\u7684\u8ddd\u96e2\u4e5f\u5c07\u53ef\u4ee5\u8a08\u7b97\u51fa\u5c0d\u61c9\u7684\u5206\u6578\u3002 \u97f3\u7d20\u5c64\u6b21\uff1a\u7576\u8a08\u7b97\u51fa\u6bcf\u500b\u97f3\u7d20\u4e2d\u56db\u9805\u8a55\u5206\u53c3\u6578\u7684\u5206\u6578\u5f8c\uff0c\u5229\u7528\u56db\u9805\u7279\u5fb5\u5c0d\u65bc\u82f1\u6587\u8a9e\u97f3\u8a55\u5206\u7cfb\u7d71\u6240\u5360\u7684\u6b0a \u91cd\u52a0\u7e3d\u5f8c\u5373\u53ef\u5f97\u5230\u6bcf\u500b\u97f3\u7d20\u7684\u5206\u6578\u3002\u4ee5\u4e0b\u662f\u8a2d\u5b9a\u7684\u516c\u5f0f\uff1a 4 3 2 1 4 3 2 1 fea fea fea fea pho score w score w score w score w score \u22c5 + \u22c5 + \u22c5 + \u22c5 = \uff0c 4 3 2 1 w w w w \u3001 \u3001 \u3001 \u5206\u5225\u4ee3\u8868\u56db\u500b\u8a55\u5206\u53c3\u6578\u7684\u6b0a\u91cd\u3002\u7d93\u7531\u4e0b\u4e00\u7bc0\u7684\u5be6\u9a57\uff0c\u6211\u5011\u53ef\u4ee5\u6c42\u51fa\u9019\u56db\u9805\u6b0a\u91cd\uff0c\u4e5f\u53ef\u4ee5\u7531\u6b0a\u91cd\u7684 \u6bd4\u4f8b\u5f97\u77e5\u56db\u9805\u8a55\u5206\u53c3\u6578\u5c0d\u65bc\u82f1\u6587\u8a55\u5206\u7684\u91cd\u8981\u6027\u3002 \u55ae\u5b57\u5c64\u6b21\uff1a\u5f97\u77e5\u6bcf\u500b\u97f3\u7d20\u7684\u5f97\u5206\u5f8c\uff0c\u4ee5\u6bcf\u500b\u97f3\u7d20\u5360\u55ae\u5b57\u7684\u6642\u9593\u70ba\u6b0a\u91cd\uff0c\u5373\u53ef\u6c42\u51fa\u53e5\u5b50\u4e2d\u6bcf\u4e00\u500b\u55ae\u5b57\u7684\u5206 \u53e5\u5b50\u5c64\u6b21\uff1a\u7531\u65bc\u55ae\u5b57\u7684\u6642\u9593\u9577\u77ed\u6703\u5f71\u97ff\u4eba\u8033\u5c0d\u65bc\u4e00\u53e5\u8a71\u7684\u95dc\u6ce8\u9ede\uff0c\u56e0\u6b64\u6211\u5011\u4e5f\u662f\u4ee5\u55ae\u5b57\u7684\u6642\u9593\u70ba\u6b0a\u91cd\u4f86 \u8a08\u7b97\u51fa\u4e00\u53e5\u8a9e\u97f3\u8a0a\u865f\u6700\u5f8c\u5f97\u5230\u7684\u5206\u6578\u3002\u4ee5\u4e0b\u70ba\u5b9a\u7fa9\u7684\u516c\u5f0f\uff1a ( ) ( ) \u2211 = \u22c5 = N n word n sen n score sentence len word len score 1 \uff0c\u5176\u4e2d N \u8868\u793a\u53e5\u5b50\u4e2d\u55ae\u5b57\u7684\u7e3d\u6578\uff0clen(x)\u8868\u793a x \u7684\u6642\u9593\u9577\u5ea6\u3002 4 \u5be6\u9a57\u7d50\u679c \u5f97\u5230\u56db\u500b\u8a55\u5206\u53c3\u6578\u4e2d\u5404\u97f3\u7d20\u7684\u5dee\u7570\u7a0b\u5ea6\u5f8c\uff0c\u6211\u5011\u4f9d\u6240\u4f54\u7684\u6bd4\u4f8b\u6c42\u51fa\u4e00\u500b\u53e5\u5b50\u7684\u5e73\u5747\u5dee\u7570\u7a0b\u5ea6\uff0c\u5373\u53ef\u4ee3\u5165\u4ee5\u4e0b\u7684 \u516c\u5f0f\uff1a ( ) ( ) ( ) ( ) 4 3 2 1 4 4 4 3 3 3 2 2 2 1 1 1 1 100 1 100 1 100 1 100 b b b b dist a w dist a w dist a w dist a w score \u22c5 + \u22c5 + \u22c5 + \u22c5 + \u22c5 + \u22c5 + \u22c5 + \u22c5 = \u5176\u4e2d 4 3 2 1 4 3 2 1 , , , , , , , b b b b a a a a \u70ba\u5dee\u7570\u7a0b\u5ea6\u8f49\u6210\u5206\u6578\u7684\u53c3\u6578\uff0c 4 3 2 1 , , , w w w w \u70ba\u56db\u500b\u8a55\u5206\u53c3\u6578\u7684\u6b0a\u91cd\uff0c\u800c , , 2 1 dist dist , 3 dist 4 dist \u8868\u793a\u6a19\u6e96\u8a9e\u97f3\u548c\u8a55\u5206\u8a9e\u97f3\u8a0a\u865f\u5728\u6bd4\u5c0d\u5f8c\u5176\u56db\u9805\u8a55\u5206\u53c3\u6578\u7684\u8ddd\u96e2\uff0c\u518d\u7d93\u7531\u4ee5\u4e0b\u7684\u5be6\u9a57\uff0c\u5373\u53ef\u6c42\u5f97\u5404\u53c3 \u6578\u503c\u3002 \u5728\u8a9e\u6599\u8a13\u7df4\u90e8\u4efd\u6211\u5011\u6536\u96c6 200 \u7d44\u8a9e\u6599\uff0c\u6bcf\u4e00\u7d44\u7684\u8a9e\u6599\u5206\u5225\u5305\u62ec\u4e00\u53e5\u6a19\u6e96\u8a9e\u97f3\u548c\u4e00\u53e5\u8a55\u5206\u8a9e\u97f3\uff0c\u6bcf\u53e5\u8a9e\u97f3 \u9577\u5ea6\u70ba 5 200 \u7d44\u8a9e\u53e5\u4f5c\u70ba\u6e2c\u8a66\u7528\u3002 \u5c07 \u9019 200 \u7d44 \u8a13 \u7df4 \u8a9e \u6599 \u900f \u904e \u8a55 \u5206 \u7cfb \u7d71 \u8a55 \u5206 \uff0c \u5247 \u6bcf \u7d44 \u8a55 \u5206 \u8a9e \u97f3 \u90fd \u6703 \u5f97 \u5230 \u56db \u500b \u7279 \u5fb5 \u5c0d \u61c9 \u7684 \u5dee \u7570 \u7a0b \u5ea6 4 3 2 1 , , , dist dist dist dist \u3002\u6536\u96c6\u4e86\u9019\u4e9b\u5dee\u7570\u7a0b\u5ea6\u548c\u5c0d\u61c9\u7684\u5206\u6578\u5f8c\uff0c\u4f7f\u7528 Simplex Downhill Search\uff0c\u5c31\u53ef\u4ee5\u627e\u51fa \u4eba\u5de5\u8a55\u5206 \u7cfb\u7d71\u8a55\u5206 Bad Average Good Bad 28 17 7 Average 20 27 20 Good 10 11 63 \u5176\u4e2d\u6a6b\u8ef8\u8868\u793a\u4eba\u5de5\u8a55\u5206\u7684\u7b49\u7d1a\u9805\u76ee\uff0c\u7e31\u8ef8\u8868\u793a\u7cfb\u7d71\u8a55\u5206\u7684\u7b49\u7d1a\u9805\u76ee\uff0c\u8868\u683c\u4e2d\u7684\u6578\u5b57\u5247\u8868\u793a\u76f8\u5c0d\u7684\u8a9e\u53e5\u6578 \u76ee\u3002\u5f9e\u8868\u4e2d\u6211\u5011\u53ef\u4ee5\u660e\u986f\u5730\u770b\u51fa\u4f86\uff0c\u5c0d\u89d2\u7dda\u7684\u6578\u76ee\u90fd\u6bd4\u540c\u4e00\u5217\u3001\u540c\u4e00\u6b04\u7684\u6578\u76ee\u9ad8\uff0c\u9019\u5c31\u8868\u793a\u5728\u7d93\u7531 Simplex Downhill Search \u8abf\u6574\u5404\u53c3\u6578\u4e4b\u5f8c\uff0c\u6211\u5011\u7684\u8a55\u5206\u7cfb\u7d71\u548c\u4eba\u5de5\u8a55\u5206\u5df2\u6709\u4e00\u5b9a\u7684\u6b63\u76f8\u95dc\u6027\uff0c\u7d04 (28+27+63) / 200 = 59%\u3002 5 \u7d50\u8ad6 \u300c\u8aaa\u8a71\u9a57\u8b49\u300d\u5c0d\u8a55\u5206\u8a9e\u97f3\u9032\u884c\u521d\u6b65\u7684\u8a55\u4f30\uff0c\u82e5\u53ef\u4fe1\u5ea6\u5920\u9ad8\uff0c\u63a5\u4e0b\u4f86\u7684\u8a55\u5206\u624d\u5177\u6709\u53ef\u4fe1\u5ea6\u3002\u300c\u8a9e\u97f3\u8a0a\u865f\u5207\u5272\u300d\u5247 \u662f\u4ee5 Forced Alignment \u5f97\u5230\u6bcf\u500b\u97f3\u7d20\u7684\u6642\u9593\u5340\u6bb5\u3002\u7d93\u7531\u5be6\u9a57\u7d50\u679c\u6211\u5011\u53ef\u4ee5\u77e5\u9053\uff0c\u4f7f\u7528\u8fa8\u8b58\u7387\u8f03\u9ad8\u7684\u8072\u5b78\u6a21\u578b\uff0c \u5176 Forced Alignment \u7684\u97f3\u7d20\u5207\u5272\u6642\u9593\u5c07\u66f4\u70ba\u6e96\u78ba\u3002\u300c\u82f1\u6587\u8a9e\u97f3\u8a55\u5206\u300d\u5305\u62ec\u8a55\u5206\u53c3\u6578\u7684\u64f7\u53d6\u3001\u5716\u6a23\u6bd4\u5c0d\u65b9\u6cd5\u7684\u8a2d \u8a08\u548c\u8a55\u5206\u6a5f\u5236\u7684\u5efa\u7acb\u7b49\u4e09\u500b\u90e8\u4efd\u3002\u85c9\u7531\u5be6\u9a57\u6211\u5011\u53ef\u4ee5\u77e5\u9053\uff0c\u300cHMM \u5c0d\u6578\u6a5f\u7387\u5dee\u7570\u300d\u5728\u82f1\u6587\u8a9e\u97f3\u8a55\u5206\u4e2d\u6240\u4ee3\u8868 \u7684\u91cd\u8981\u6027\u6700\u9ad8\uff0c\u800c\u300c\u97f3\u91cf\u5f37\u5ea6\u66f2\u7dda\u300d\u5247\u662f\u6700\u4f4e\u3002 \u95dc\u65bc\u5be6\u9a57\u6e2c\u8a66\u8a9e\u6599\u7684\u90e8\u4efd\uff0c\u6211\u5011\u4f7f\u7528\u4e86 1,680 \u53e5\u6bcd\u8a9e\u70ba\u82f1\u6587\u7684\u8a9e\u97f3\u6a94\u6848\uff0c\u5176\u8a9e\u6599\u7684\u9577\u5ea6\u7e3d\u548c\u7d04\u70ba 1 \u53c9\u5be6\u9a57\u3002\u8868\u683c 2 \u5217\u51fa\u97f3\u7d20\u5207\u5272\u6b63\u78ba\u7387\u7684\u5be6\u9a57\u7d50\u679c\uff1a \u8868\u683c 3 \u82f1\u6587\u8a9e\u97f3\u8fa8\u8b58\u7387 \u5be6\u9a57\u65b9\u5f0f \u9805\u76ee N-Wave /N-HMM N-Wave /T-HMM T-Wave /N-HMM T-Wave /T-HMM \u70ba\u4e86\u8a08\u7b97 HMM \u5c0d\u6578\u6a5f\u7387\u7684\u5dee\u7570\uff0c\u6211\u5011\u5148\u4ee5 N-HMM(HMM trained from Native Speaker)\u6c42\u51fa\u6a19\u6e96\u8a9e\u97f3\u8a0a\u865f \u53ca\u8a55\u5206\u8a9e\u97f3\u8a0a\u865f\u4e2d\u6bcf\u500b\u97f3\u7d20\u7684\u5c0d\u6578\u6a5f\u7387\uff0c\u82e5\u5c0d\u6578\u6a5f\u7387\u503c\u6108\u5927\uff0c\u8868\u793a\u8a72\u97f3\u7d20\u7684\u767c\u97f3\u6108\u63a5\u8fd1\u8072\u5b78\u6a21\u578b\u3002\u5716\u8868 9 \u70ba HMM \u5c0d\u6578\u6a5f\u7387\u5dee\u7570\u6bd4\u5c0d\u7684\u6d41\u7a0b\u5716\uff1a \u6578\uff0c\u4ee5\u4e0b\u70ba\u8a2d\u5b9a\u7684\u516c\u5f0f\uff1a ( \u8a9e\u97f3\u8a55\u5206\u7684\u904b\u7528\u76f8\u7576\u5ee3\u6cdb\u4e14\u5be6\u7528\uff0c\u914d\u5408\u672a\u4f86\u6280\u8853\u7684\u6210\u719f\uff0c\u4e0d\u53ea\u53ef\u4f5c\u70ba\u82f1\u8a9e\u5b78\u7fd2\u7684\u5de5\u5177\uff0c\u4e4b\u5f8c\u7684\u53f0\u8a9e\u3001\u5ba2 ) ( ) \u2211 = \u22c5 = N n pho n word n score word len pho len \u8a9e\u8a55\u5206\u5b78\u7fd2\u4e5f\u5c07\u662f\u53f0\u7063\u5730\u5340\u91cd\u8981\u7684\u7814\u7a76\u4e4b\u4e00\u3002 score 1 \uff0c\u5176\u4e2d N \u70ba\u4e00\u55ae\u5b57\u4e2d\u8a55\u5206\u97f3\u7d20\u7684\u6578\u91cf\uff0clen(x)\u8868\u793a x \u7684\u6642\u9593\u9577\u5ea6\u3002 \u53c3\u8003\u8cc7\u6599</td></tr><tr><td>2 \u7d93\u7531\u4e0a\u8ff0\u7684\u5be6\u9a57\uff0c\u6211\u5011\u5f97\u5230\u97f3\u91cf\u5f37\u5ea6\u66f2\u7dda\u7684\u6b0a\u91cd\u70ba 7.45%\uff0c\u57fa\u983b\u8ecc\u8de1\u66f2\u7dda\u7684\u6b0a\u91cd\u70ba 22.40%\uff0c\u767c\u8072\u6025\u7de9\u8b8a\u5316 2 ) ( ) ( Evaul stard P Const Const factor + = \u7684\u6b0a\u91cd\u70ba 17.24%\uff0cHMM \u5c0d\u6578\u6a5f\u7387\u5dee\u7570\u7684\u6b0a\u91cd\u70ba 52.91%\u3002</td></tr><tr><td>\u7576\u7b97\u51fa\u6a19\u6e96\u8a9e\u97f3\u548c\u8a55\u5206\u8a9e\u97f3\u7684 Const \u503c\u5f8c\uff0c\u518d\u7d93\u7531\u5e73\u65b9\u76f8\u52a0\u5373\u53ef\u5f97\u5230\u6a5f\u7387\u500d\u6578 \u63a5\u8457\u6211\u5011\u5c07 200 \u53e5\u6e2c\u8a66\u8a9e\u53e5\u7684\u4eba\u5de5\u8a55\u5206\u7d50\u679c\u5206\u6210\u4e09\u500b\u7b49\u7d1a\uff1aBad(0~59)\u3001Average(60~79)\u3001Good(80~100)\uff0c P factor \uff0c\u5c07\u6b64\u6a5f\u7387\u500d\u6578\u4e58 \u53e6\u5916\u4e5f\u628a\u5c07 200 \u53e5\u6e2c\u8a66\u8a9e\u53e5\u7684\u7cfb\u7d71\u8a55\u5206\u7d50\u679c\u4f9d\u6b64\u5206\u6210\u4e09\u500b\u7b49\u7d1a\u3002\u6700\u5f8c\u518d\u7d71\u8a08\u6bcf\u500b\u53e5\u5b50\u7684\u4eba\u5de5\u8a55\u5206\u548c\u7cfb\u7d71\u8a55\u5206 \u4e0a\u5169\u8a9e\u97f3\u8a0a\u865f\u5c0d\u6578\u6a5f\u7387\u7684\u5dee\u8ddd\u5c31\u662f\u6211\u5011\u767c\u97f3\u7279\u5fb5\u7684\u5dee\u7570\u7a0b\u5ea6\u3002 \u5f8c\uff0c\u5c31\u53ef\u4ee5\u5f97\u5230\u8868\u683c 5\u7684\u7d50\u679c\uff1a</td></tr><tr><td>\u5716\u8868 8 \u8a55\u5206\u7cfb\u7d71\u6d41\u7a0b\u5716</td></tr></table>",
                "html": null,
                "type_str": "table",
                "text": "\u5c0f\u6642 23 \u5206 51 \u79d2\uff0c\u4ee5\u4e0b\u6211\u5011\u7c21\u7a31\u70ba N-Wave (Waves from Native-Speaker)\u3002\u53e6\u5916\u4f7f\u7528\u4e86 2,342 \u53e5\u6bcd\u8a9e\u70ba\u570b\u8a9e\u7684\u8a9e\u97f3\u6a94 \u6848\uff0c\u8a9e\u6599\u7684\u9577\u5ea6\u7e3d\u548c\u7d04\u70ba 1 \u5c0f\u6642 57 \u5206 43 \u79d2\uff0c\u4ee5\u4e0b\u7c21\u7a31\u70ba T-Wave(Waves from Taiwanese)\uff0c\u4f86\u505a Outside Test\u3002 \u5be6\u9a57\u7528\u7684\u8a9e\u6599\u5176\u97f3\u8a0a\u683c\u5f0f\u7686\u70ba PCM\uff0c\u97f3\u8a0a\u53d6\u6a23\u983b\u7387\u70ba 16 kHz\uff0c\u4f4d\u5143\u89e3\u6790\u5ea6\u70ba 16 bits\u3002 \u5728\u8072\u5b78\u6a21\u578b\u9019\u500b\u90e8\u4efd\uff0c\u6211\u5011\u8a13\u7df4\u51fa\u4e86\u5169\u500b\u8072\u5b78\u6a21\u578b\uff1a\u4e00\u500b\u662f\u7531\u4ee5\u82f1\u6587\u4f5c\u70ba\u6bcd\u8a9e\u7684\u4f7f\u7528\u8005\u6240\u9304\u88fd\u7684\u8a13\u7df4\u8a9e\u6599 \u7522\u751f\u7684\u8072\u5b78\u6a21\u578b\uff0c\u4ee5\u4e0b\u6211\u5011\u7c21\u7a31\u70ba N-HMM(HMM trained from Native-Speaker)\uff0c\u53e6\u4e00\u500b\u5247\u662f\u7531\u81fa\u7063\u4eba\u6240\u9304\u88fd\u7684 \u8a13\u7df4\u8a9e\u6599\u6240\u7522\u751f\u7684\uff0c\u4ee5\u4e0b\u6211\u5011\u7c21\u7a31\u70ba T-HMM(HMM trained from Taiwanese)\u3002 \u95dc\u65bc\u5be6\u9a57\u7684\u65b9\u5f0f\uff0c\u6211\u5011\u5206\u5225\u5c0d\u6bcf\u4e00\u53e5\u8a9e\u97f3\u8a0a\u865f\u548c\u5df2\u77e5\u7684\u8a9e\u97f3\u5167\u5bb9\u6587\u5b57\u4f5c Forced Alignment\uff0c\u518d\u7531\u7522\u751f\u7684\u7d50 \u679c\u5c0d\u6bcf\u500b\u55ae\u5b57\u53ca\u97f3\u7d20\u5224\u65b7\u5176\u6642\u9593\u5340\u6bb5\u7684\u5207\u5272\u662f\u5426\u6b63\u78ba\u3002 \u70ba\u4e86\u6bd4\u8f03\u5169\u500b\u8072\u5b78\u6a21\u578b\u6240\u7522\u751f\u7684\u5f71\u97ff\uff0c\u6211\u5011\u5c0d\u8a9e\u6599(N-Wave, T-Wave) \u548c\u8072\u5b78\u6a21\u578b(N-HMM, T-HMM)\u4f5c\u4ea4 \u79d2\u3001\u97f3\u8a0a\u683c\u5f0f\u70ba PCM\u3001\u97f3\u8a0a\u53d6\u6a23\u983b\u7387\u70ba 16 kHz\u3001\u4f4d\u5143\u89e3\u6790\u5ea6\u70ba 16 bits\u3002\u5176\u4e2d\u6a19\u6e96\u8a9e\u97f3\u7684\u8a9e\u6599\u9577\u5ea6\u7e3d \u548c\u7d04\u70ba 12 \u5206 51 \u79d2\uff0c\u8a55\u5206\u8a9e\u97f3\u7684\u8a9e\u6599\u9577\u5ea6\u7e3d\u548c\u7d04\u70ba 18 \u5206 39 \u79d2\u3002\u63a5\u8457\u8acb\u5916\u8a9e\u6240\u8001\u5e2b\u5354\u52a9\u6211\u5011\u5c0d\u6bcf\u4e00\u53e5\u8a55\u5206\u8a9e\u97f3 \u4f5c\u4e3b\u89c0\u7684\u8a55\u5206\uff0c\u4e4b\u5f8c\u518d\u7d71\u8a08\u5be6\u9a57\u4e2d\u6bcf\u4e00\u53e5\u8a9e\u97f3\u4eba\u70ba\u8a55\u5206\u7684\u5e73\u5747\u5206\u6578\u3002\u540c\u6a23\u7684\uff0c\u6309\u7167\u8a13\u7df4\u8a9e\u53e5\u7684\u4f5c\u6cd5\uff0c\u6211\u5011\u4e5f\u6536 \u96c6\u4e86",
                "num": null
            }
        }
    }
}