约束条件下的结构化高斯混合模型及非平行语料语音转换

doi:10.3969/j.issn.0372-2112.2016.09.037

电子学报 ›› 2016, Vol. 44 ›› Issue (9): 2282-2288.DOI: 10.3969/j.issn.0372-2112.2016.09.037

约束条件下的结构化高斯混合模型及非平行语料语音转换

车滢霞, 俞一彪

苏州大学电子信息学院, 江苏苏州 215006

收稿日期:2015-02-08 修回日期:2015-08-10 出版日期:2016-09-25
- 通讯作者:
- 俞一彪
- 作者简介:
- 车滢霞 女,1989年生,江苏常州人,苏州大学电子信息学院硕士,研究方向为语音信号处理.
- 基金资助:
- 国家自然科学基金 (No.61271360）; 江苏省自然科学基金 (No.BK20131196）

Non-parallel Corpora Voice Conversion Based on Structured Gaussian Mixture Model Under Constraint Conditions

CHE Ying-xia, YU Yi-biao

School of Electronic and Information Engineering, Soochow University, Suzhou, Jiangsu 215006, China

Received:2015-02-08 Revised:2015-08-10 Online:2016-09-25 Published:2016-09-25
- Supported by:
- National Natural Science Foundation of China (No.61271360); Natural Science Foundation of Jiangsu Province, China (No.BK20131196)

摘要/Abstract

摘要：

提出一种约束条件下的结构化高斯混合模型及非平行语料语音转换方法.从源与目标说话人的原始非平行语料中提取出少量相同音节，在结构化高斯混合模型的训练过程中，利用这些相同音节包含的语义信息及声学特征对应关系对K均值聚类中心进行约束，并在（Expectation Maximum，EM）迭代过程中对语音帧属于模型分量的后验概率进行修正，得到基于约束的结构化高斯混合模型（Structured Gaussian Mixture Model with Constraint condition，C-SGMM）.再利用全局声学结构（Acoustic Universal Structure，AUS）原理对源和目标说话人的约束结构化高斯混合模型的高斯分布进行匹配对准，推导出短时谱转换函数.主观和客观评价实验结果表明，使用该方法得到的转换后语音在谱失真，目标倾向性和语音质量等方面均优于传统的结构化模型语音转换方法，转换语音的平均谱失真仅为0.52，说话人正确识别率达到95.25%，目标语音倾向性指标ABX平均为0.82，性能更加接近于基于平行语料的语音转换方法.

关键词: 语音转换, 结构化高斯混合模型, 非平行语料, 约束条件

Abstract:

This paper proposes a structured Gaussian mixture model with constraint conditions (C-SGMM) for non-parallel corpora voice conversion.A small number of voice signals with the same syllables from the source and target non-parallel corpus are extracted as constraint conditions,then the correspondence between acoustic features of source and target corpus formed by these syllables are applied in the process of statistical acoustic model training.The constraint conditions are used to restrict the cluster centers of K-means clustering process,and they are also used in EM algorithm to adjust the voice frame's posterior probability belonging to a Gaussian distribution component for model training.Then Gaussian distributions in source and target structured Gaussian mixture models are aligned using acoustic universal structure principle and the conversion function can be derived.Results of both subjective and objective experiments indicate that the conversion performance obtained by the proposed method are advanced to that of the traditional structured method in cepstrum distortion,target tendency and speech quality aspects.The average cepstrum distortion of converted speech is only 0.52,the speaker recognition rate of the converted speech reaches 95.25%,and the performance closer to the conventional parallel corpora GMM based method is achieved.

Key words: voice conversion, structure Gaussian mixture model, non-parallel corpora, constraint conditions

中图分类号:

TN912.33

车滢霞, 俞一彪. 约束条件下的结构化高斯混合模型及非平行语料语音转换[J]. 电子学报, 2016, 44(9): 2282-2288.

CHE Ying-xia, YU Yi-biao . Non-parallel Corpora Voice Conversion Based on Structured Gaussian Mixture Model Under Constraint Conditions[J]. Acta Electronica Sinica, 2016, 44(9): 2282-2288.

参考文献

[1] Stylianou,Y.Voice transformation:A survey[A].IEEE International Conference on Acoustics,Speech and Signal Processing[C].Taipei:IEEE,2009.3585-3588.
[2] 康永国,双志伟,陶建华,张维.基于混合映射模型的语音转换算法研究[J].声学学报,2006,31(6):555-562. Kang Yongguo,Shuang Zhiwei,Tao Jianhua,et al.A hybrid method to convert acoustic features for voice conversion[J].Acta Acustica,2006,31(6):555-562.(in Chinese)
[3] 徐宁,杨震,张玲华.基于状态空间模型的子频带语音转换算法[J].电子学报,2010,38(3):646-653. Xu Ning,Yang Zhen,Zhang Ling-hua.Sub-and voice morphing algorithm based on state-space model[J].Acta Electronica Sinica,2010,38(3):646-653.(in Chinese)
[4] Gu Hung-yan,Tsai Sung-fung.Improving segmental GMM based voice conversion method with target frame selection[A].International Symposium on Chinese Spoken Language Processing (ISCSLP)[C].Singapore:IEEE,2014.483-487.
[5] Mouchtaris A,Van der Spiegel J,Mueller P.Nonparallel training for voice conversion based on a parameter adaptation approach[J].IEEE Transactions on Audio,Speech,and Language Processing,2006,14(3):952-963.
[6] Erro D,Moreno A,Bonafonte A.INCA Algorithm for training voice conversion systems from nonparallel corpora[J].IEEE Transactions on Audio,Speech,and Language Processing,2010,18(5):944-953.
[7] Saito D,Watanabe S,Nakamura A,et al.Statistical voice conversion based on noisy channel model[J].IEEE Transactions on Audio,Speech,and Language Processing,2012,20(6):1784-1794.
[8] 俞一彪,曾道建,姜莹.采用独立说话人模型的语音转换[J].声学学报,2012,37(3):346-352. Yu Yibiao,Zheng Daojiang,Jiang Ying.Voice conversion based on isolated speaker model[J].Acta Acustica,2012,37(3):346-352.(in Chinese)
[9] Li Xian,Wang Zeng-fu.Frame correlation based autoregressive GMM method for voice conversion[A].International Symposium on Chinese Spoken Language Processing (ISCSLP)[C].Singapore:IEEE,2014.221-225.
[10] Li Yan-ping,Zhang Ling-hua,Ding Hui.Nonparallel voice conversion based on phoneme classification and eigenvoices[A].IEEE International Conference on Communication Technology (ICCT)[C].Nanjing:IEEE,2010.662-665.
[11] Oliva G,La Manna D,Fagiolini A,et al.Distance-constrained data clustering by combined k-means algorithms and opinion dynamics filters[A].Mediterranean Conference of Control and Automation (MED)[C].Palermo:IEEE,2014.612-619.
[12] 於跃成.基于半监督学习的分布式和演化聚类研究[D].南京:南京航空航天大学,2012. Yu Yuecheng.Distributed clustering and evolutionary clustering algorithm based on semi-supervised learning[D].Najing:Nanjing University of Aeronautics and Astronautics,2012.(in Chinese)
[13] Michael Pitz,Hermann Ney.Vocal tract normalization equals linear transformation in cepstral space[J].IEEE Trans on Audio,Speech,and Language Processing,2005,13(5):930-944.
[14] Minematsu N.Mathematical evidence of the acoustic universal structure in speech[A].IEEE International Conference on Acoustics,Speech,and Signal Processing,(ICASSP)[C].Philadelphia:IEEE,2005.889-892.
[15] Che Yingxia,Yu Yibiao.Fast matching algorithm between statistical acoustic models of source-target speaker in structured approach of non-parallel corpora voice conversion[A].IEEE International Conference on Information Science and Technology (ICIST)[C].Shenzhen:IEEE,2014.88-92.

约束条件下的结构化高斯混合模型及非平行语料语音转换

Non-parallel Corpora Voice Conversion Based on Structured Gaussian Mixture Model Under Constraint Conditions

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 5

编辑推荐

Metrics

[1]	钱兆鹏, 肖克晶, 刘蝉, 孙悦. 基于语义的汉语普通话电子喉语音转换增强[J]. 电子学报, 2020, 48(5): 840-845.
[2]	徐宁;杨震;张玲华. 基于状态空间模型的子频带语音转换算法[J]. 电子学报, 2010, 38(3): 646-653.
[3]	刘政林;邹雪城;向祖权;肖建平;赵慧波;李仕杰. 定标器的设计与实现[J]. 电子学报, 2006, 34(1): 185-188.
[4]	汤炜;李清亮;焦培南;吴振森. 无时间约束FDTD方法在三维散射中的应用[J]. 电子学报, 2004, 32(9): 1499-1503.
[5]	方振贤;汪鹏君;刘莹. 二值、多值和绝热电路通用的电路理论[J]. 电子学报, 2003, 31(2): 303-305.