Research on the Application of Variational Autoencoder in Environmental Noise Cancellation
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 L2 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.
NETL
NSTL
万方数据
