基于深度学习的广播电视录音技术与声音转换算法研究  

Research on Broadcasting and Television Recording Technology Assisted by Deep Learning and Sound Conversion Algorithm

作　　者：白泉 BAI Quan(Henan Broadcasting System,Zhengzhou 450008,China)

机构地区：[1]河南广播电视台,河南郑州450008

出　　处：《电声技术》2025年第1期100-102,共3页Audio Engineering

摘　　要：深入研究深度学习辅助的音频增强、噪声去除、语音识别等技术,并建立基于卷积神经网络、循环神经网络以及生成对抗网络的声音转换模型。通过大量音频数据的训练和迁移检验,给出模型在梅尔倒谱失真、主观语音质量评估及平均意见得分等评估指标上的表现。研究表明,深度学习辅助可进一步提升录音技术的性能,为广播电视节目制作提供技术支持。In-depth study of deep learning assisted audio enhancement,noise removal,speech recognition and other technologies,and the establishment of voice conversion models based on convolutional neural networks,circular neural networks and generating confrontation networks.Through the training and migration test of a large number of audio data,the performance of the model in Mel cepstrum distortion,subjective speech quality evaluation and average opinion score is given.The research shows that deep learning can further improve the performance of recording technology and provide technical support for radio and television program production.

关 键 词：深度学习辅助 广播电视录音 声音转换算法 

分 类 号：TN912.3[电子电信—通信与信息系统]