摘要
针对语音信号中的环境噪声影响问题,研究一种基于变分自编码器(Variational Auto-Encoder,VAE)的噪声消除方法,并进行了相关实验验证。首先,深入研究VAE的基本原理;其次,引入L_2正则化方法优化VAE,以提高模型的泛化能力和健壮性;最后,利用ESC-50和VCTK数据集构建包含环境噪声的语音信号数据集,并在此基础上进行了一系列实验。实验结果表明,所提方法能够有效降低环境噪声对语音信号的影响。
This article focuses on the impact of environmental noise in speech signals,and studies a noise cancellation method based on Variational Auto-Encoder(VAE),and conducts relevant experimental verification.Firstly,this article delves into the basic principles of VAE;Subsequently,the L,regularization method was introduced to optimize the VAE,in order to improve the model's generalization ability and robustness;Subsequently,a speech signal dataset containing environmental noise was constructed using the ESC-50 and VCTK datasets,and a series of experiments were conducted on this basis.The experimental results show that the proposed method can effectively reduce the impact of environmental noise on speech signals.
作者
李红玲
LI Hongling(Guangxi Electronic Senior Technical School,Nanning 530031,China)
出处
《电声技术》
2024年第8期105-107,共3页
Audio Engineering
关键词
变分自编码器(VAE)
正则化
环境噪声
去噪
Variational Auto-Encoder(VAE)
regularization
environmental noise
denoising
