利索-全球专利检索

  1. 登录
  2. 注册
发明公开
CN107221321A 一种用于任意源和目标语音之间的语音转换方法 无效 - 驳回

一种用于任意源和目标语音之间的语音转换方法
摘要：
本发明公开了一种针对任意源和目标语音之间的语音转换方法，包括以下步骤，步骤一，建立至少一个目标语音的张量字典；步骤二，构建目标语音对应的语音字典和任意源的语音字典；步骤三，重构任意源的语音内容，实现任意源到目标语音的转换。本发明利用在同一张张量字典中，利用任意源、与要转换的目标语音对应张量字典中语音信号，构造出各自的语音字典，实现语音转换；无需每次语音转换之前搜集到大量的任意源和目标说话人的平行语音进行训练，实用性及语音转化效率更高。
摘要（英）：

The invention discloses a speech conversion method between any sources and a target speech. The method comprises the following steps of step1, establishing a tensor dictionary of at least one target speech; step2, constructing a speech dictionary corresponding to the target speech and a speech dictionary of any sources; and step3, reconstructing a speech content of the any sources and realizing conversion from the any sources to the target speech. In a same tensor dictionary, the any sources and a speech signal in the tensor dictionary corresponding to the target speech to be converted are used to construct a respective speech dictionary and realize speech conversion. A condition that a lot of any sources and a parallel speech of a target speaker are collected so as to carry out training before each speech conversion is not needed, and practicality and speech conversion efficiency are high.

0/0