权益分类	功能权益	普通用户	{{item.name}}会员
{{category.name}}	{{benefitItem.name}}

Development of Speech Synthesizing Device with individuality Features by LSP parameter

利用LSP参数开发具有个性特征的语音合成装置

基本信息

批准号：
03452170
负责人：
SUGATA Kazuhiro
金额：
$ 1.92万
依托单位：
Tottori University
依托单位国家：
日本
项目类别：
Grant-in-Aid for General Scientific Research (B)
财政年份：
1991
资助国家：
日本
起止时间：
1991 至 1993
项目状态：
已结题

来源：
https://kaken.nii.ac.jp/en/grant/KAKENHI-PROJECT-03452170/
关键词：
LSP parameters speech synthesis voice source individuality feature voice conversion connecting parameters glide neural network

项目摘要

In our research on development of speech synthesizing device awarded with individuality features by using LSP parameters, we placed focus on the following three subjects. (1) a new approach of constructing voice source for speech synthesis of high quality. (2) effective methods for connecting properly the parameters, such as LSP, corresponding to phonemic components in order to generate continuous speech signal. (3) voice conversion method between male and female using LSP parameters analysis and synthesis.When we are required to provide the individuality features to the synthesized voice, it is essential to generate good quality voice source. Inferior quality voice source causes the deterioration in the synthesized voice. Thefirst subject above was studied elaborately in order to obtain desirable voice source. We proposed a method to generate voice source shich is particularly effective to synthesize the clear consonant voices. The second subject was done to develop the method in speech synthesis to connect the phonemic components which are stored in the device. If we can connect the phonemic parameter components properly, we can synthesize good quality voices. We proposed a very effective method by using neural network. In the third subject above, we proposed a speech synthesis method to control pitch of voice which does no cause the degradation of individuality features. Furthermore, we porposed a method to control the voice properties.

在利用LSP参数开发具有个性化特征的语音合成器的研究中，我们重点研究了以下三个方面的内容。(1)为实现高质量语音合成，提出了一种新的声源构造方法。(2)正确连接与音素成分对应的参数，如LSP，以产生连续的语音信号的有效方法。(3)基于LSP参数分析和合成的男女语音转换方法，当需要对合成的语音提供个性化特征时，必须生成高质量的声源。劣质声源会导致合成语音的质量恶化。为了获得理想的声源，对上述第一个课题进行了细致的研究。我们提出了一种语音源的生成方法，该方法对合成清晰的辅音特别有效。第二个主题是开发语音合成中连接存储在设备中的音素分量的方法。如果能够正确地连接音素参数分量，就可以合成出高质量的声音。我们提出了一种非常有效的神经网络方法。在上面的第三个主题中，我们提出了一种语音合成方法来控制语音的基音高，而不会导致个性特征的退化。此外，我们还提出了一种控制语音属性的方法。