【24h】

Neutral to Target Speech Conversion using Polynomial Curve Fitting.

机译:使用多项式曲线拟合进行中性到目标语音转换。

获取原文
获取原文并翻译 | 示例

摘要

This paper propose a work in emotion conversion of neutral speech signal to target signals like sad speech, fear speech and anger speech signals. A method called curve fit using polynomials is proposed for converting the given neutral speech signal to a target emotional speech signals like sad and fear. As pitch plays an important role in emotion expression, here pitch contour is modified first by using the method of curve fit using polynomials and then speech parameters are modified. Sad and fear speech signals are synthesized by modifying speech/prosodic parameters like pitch contour, intensity contour and duration contour. Anger speech signal are synthesized by modifying prosodic parameters like pitch range, shift pitch frequency, intensity contour and duration contour. The target speech is compared with resultant speech obtained by modifying only the pitch contour and also have observed the improvements in quality of expression due to modification of spectral energy. The method curve fit using polynomials results in good quality for neutral to sad speech and neutral to fear speech conversions and from neutral to anger the speech conversion quality is not up to the mark.
机译:本文提出了一种将中性语音信号情感转换为目标信号(如悲伤语音,恐惧语音和愤怒语音信号)的工作。提出了一种使用多项式的曲线拟合方法,用于将给定的中性语音信号转换为目标情感语音信号,例如悲伤和恐惧。由于音调在情感表达中起着重要的作用,因此这里的音调轮廓首先通过使用多项式的曲线拟合方法进行修改,然后修改语音参数。悲伤和恐惧的语音信号是通过修改语音/韵律参数(如音高轮廓,强度轮廓和持续时间轮廓)来合成的。愤怒的语音信号是通过修改韵律参数(如音调范围,移位音调频率,强度轮廓和持续时间轮廓)来合成的。将目标语音与仅通过修改音高轮廓而获得的最终语音进行比较,并且还观察到由于频谱能量的修改而导致的表达质量的提高。使用多项式的方法曲线拟合可为中性到悲伤的语音和中性到恐惧的语音转换提供良好的质量,从中性到愤怒的语音转换质量达不到标准。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号