基于深度学习的广播音频输入信号噪声处理方法  

Noise Processing Method of Broadcast Audio Input Signal Based on Deep Learning

作　　者：李波 LI Bo(Chongqing Broadcasting and Television Technology Center 529 Units,Chongqing 409000,China)

机构地区：[1]重庆广播电视技术中心529台,重庆409000

出　　处：《电声技术》2024年第11期82-84,94,共4页Audio Engineering

摘　　要：提出一种基于深度学习的广播音频噪声处理方法,旨在提高广播音频质量。通过采用卷积神经网络进行噪声分类与定位,并结合深度神经网络(Deep Neural Network,DNN)与最小均方误差(Minimum Mean Squared Error,MMSE)滤波器实现自适应噪声抑制。实验结果表明,所提方法在信号失真比和感知评价语音质量指标上显著优于传统方法,展现出卓越的降噪效果。This paper proposes a method for processing broadcast audio noise based on deep learning,aiming at improving broadcast audio quality.The convolutional neural network is used for noise classification and location,and the adaptive noise suppression is realized by combining the Deep Neural Network(DNN)and the Minimum Mean Squared Error(MMSE)filter.The experimental results show that the proposed method is significantly superior to the traditional methods in signal distortion ratio and perceptual evaluation of speech quality,showing excellent noise reduction effect.

关 键 词：深度学习 广播音频 噪声处理 自适应噪声抑制 

分 类 号：TN93[电子电信—信号与信息处理]