摘要
文章基于星型生成式对抗网络-语音转换(Star Generative Adversarial Networks-Voice Conversion,StarGAN-VC)模型,研究了一种先进的语音风格转换技术,旨在实现对语音信号的高效转换。首先,详细阐述了基于StarGAN-VC的语音转换方法的基本原理。其次,深入研究特征提取和基频转换方法,以及StarGAN-VC模型的数学原理。最后,通过在VCC2018数据集上的实验,验证了该方法的性能。实验结果表明,该方法在频谱包络相似度和基频准确度等指标上均取得了令人满意的效果。
Based on the Star Generative Adversary Networks-Voice Conversion(StarGAN-VC)model,this paper studies an advanced voice style conversion technology,aiming at achieving efficient conversion of voice signals.Firstly,the basic principle of voice conversion method based on StarGAN-VC is expounded in detail.Secondly,the methods of feature extraction and fundamental frequency conversion and the mathematical principle of StarGAN-VC model are deeply studied.Finally,the performance of this method is verified by experiments on VCC2018 data sets.The experimental results show that this method has achieved satisfactory results in spectral envelope similarity and fundamental frequency accuracy.
作者
申少鹏
胡松涛
SHEN Shaopeng;HU Songtao(Henan Vocational and Technical College of Architecture,Zhengzhou 450000,China)
出处
《电声技术》
2024年第1期35-37,共3页
Audio Engineering
作者简介
申少鹏(1985-),男,硕士,助教,研究方向为计算机科学与技术。
