| 题名 | 汉语言语数据库自动标注系统的研究 |
| 作者 | 朱维彬 |
| 学位类别 | 博士 |
| 答辩日期 | 1998 |
| 授予单位 | 中国科学院中科院声学研究所 |
| 授予地点 | 中科院声学研究所 |
| 关键词 | 言法数据库 自动标准 HMM模型 音子 音段边界 |
| 中文摘要 | 明确了汉语语音学单元为声、介、韵（调）。确定了汉语言语数据库的语音学标注层次为宽式音标标注，编纂了汉语音标键盘符号系统——汉语SAMPA-X。对孤立音节、试验句中的音段边界的声学线索，手工切分所依据的准则和采取的策略进行了总结，对连续语音中的音变现象进行了总结。分析手工切分不稳定性产生的原因。证实了可以按四呼对声母声学体现加以分类；确定了声母声学建模所对应的音段范围。分析了语速对音段时长变化的影响、语速对语调的影响。提出了基出了基于HMM的汉语语音自动标注系统统计学模型，在正则音标系统中采用了右语境有关的模型以体现协同发音的影响；在宽式音标系统中，设计了包括各种可能音位变体的发音模型。提出了基于MFCC相关系数的参数系统和HMM系统相互校验的自动标注系统，用以解决切分结果可靠性的自动判决问题。 |
| 英文摘要 | The initial, medial and final are confirmed as the phonetic units (e.g. phones) for Chinese speech. The labeling done is defined at the broad-phonetic level, and a computer-coding IPA symbol system is refined to annotate the speech segment. The acoustic cues for detecting the phone boundary are pursued, and the method used for segmentation and labeling is determined, and the allophones in utterance are investigated also. The re-labeled utterances are used to pursue the cases, which distribute the inconsistency of judgment in manual segmentation. It is confirmed that the acoustic features of the initial can be divided into four clusters according to the status of the medial. The region of segment for modeling an initial is determined. The interaction between speech rate and segment duration is investigated. And the interaction between speech rate and intonation is investigated also. A statistic model based on HMM is chosen to construct the system of automatic phonetic labeling for chinese speech. In the orthographic labeling system, the right context-dependent model is adopted to deal with the influence of co-articulation between two adjacent phones, the state number of HMM model is defined respectively according to the manner of articulation of the phone. In the broad-phonetic labeling system, a pronunciation model based on phonetic knowledge of Chinese speech is formed. As addition means, the correlation coefficient of MFCC is used to detect the reliability of the output from HMM system.(图版 57个; 表格 33个; 参考文献 121个) |
| 语种 | 中文 |
| 公开日期 | 2011-05-07 |
| 页码 | 108 |
| 内容类型 | 学位论文 |
| 源URL | [http://159.226.59.140/handle/311008/1418]  |
| 专题 | 声学研究所_声学所博硕士学位论文_1981-2009博硕士学位论文 |
| 推荐引用方式 GB/T 7714 | 朱维彬. 汉语言语数据库自动标注系统的研究[D]. 中科院声学研究所. 中国科学院中科院声学研究所. 1998. |
| 个性服务 |
| 查看访问统计 |
| 相关权益政策 |
| 暂无数据 |
| 收藏/分享 |
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。
修改评论