基于神经网络的高保真音频重建方法研究  

Research on High Fidelity Audio Reconstruction Method Based on Neural Network

作　　者：罗明挽 LUO Mingwan(Yangjiang Vocational and Technical College,Yangjiang 529500,China)

机构地区：[1]阳江职业技术学院,广东阳江529500

出　　处：《电声技术》2024年第10期86-88,共3页Audio Engineering

摘　　要：为提升受损音频信号的质量,提出一种基于神经网络的高保真音频重建方法。该方法利用卷积神经网络(Convolutional Neural Network,CNN)提取特征,利用自适应滤波器抑制噪声,利用生成对抗网络进行音频重建。实验结果表明,在不同噪声环境下,该方法能显著提高音频质量,尤其在低信噪比(Signal-to-Noise Ratio,SNR)条件下表现优异。To improve the quality of damaged audio signals,a high fidelity audio reconstruction method based on neural networks is proposed.This method utilizes Convolutional Neural Network(CNN)to extract features,adaptive filters to suppress noise,and Generative Adversarial Network for audio reconstruction.The experimental results show that this method can significantly improve audio quality in different noise environments,especially under low Signal-to-Noise Ratio(SNR)conditions.

关 键 词：高保真音频 重建 卷积神经网络(CNN) 

分 类 号：TS801.3[轻工技术与工程]