深度学习理论及其应用专题讲座(四) 第8讲 深度学习方法在语音增强中的应用  

Deep Learning Based Speech Enhancement Technology and Method

作　　者：时文华[1] 张雄伟[2] 张瑞昕 韩伟[4] SHI Wen-hua ZHANG Xiong-wei ZHANG Rui-xin HAN Wei(Postgraduate Team 1 CCIS, PLAUST, Nanjing 210007, China College of Command Information Systems, PLAUST PLA Institute of National Defense Information, Wuhan 430000, China Postgraduate Team 2 CCIS, PLAUST)

机构地区：[1]解放军理工大学指挥信息系统学院研究生1队,江苏南京210007 [2]解放军理工大学指挥信息系统学院 [3]国防信息学院,湖北武汉430000 [4]解放军理工大学指挥信息系统学院研究生2队

出　　处：《军事通信技术》2016年第3期98-104,共7页Journal of Military Communications Technology

基　　金：国家自然科学基金资助项目(61402519;61471394);江苏省自然科学基金资助项目(BK20140071;BK20140074;BK2012510)

摘　　要：语音增强旨在提高受损语音的质量和可懂度,在语音通信、语音编码和语音识别等诸多领域得到广泛的应用。近年来深度学习依靠其良好的复杂特征提取表达能力,擅长对数据中的结构化信息进行建模能力,在语音信号和信息处理研究领域受到重视并取得成功的应用。文章简要介绍单通道语音增强方法的分类、常用模型,探讨深度学习方法在语音增强中的应用,对语音增强技术中常见的深度网络模型进行了介绍,最后对应用前景进行了展望。To improve the quality and intelligibility, speech enhancement has been widely used in speech communication, speech coding and speech recognition. Depending on its higher ability to extract complex features and model structured information in data, deep learning has emerged as a powerful machine learning approach. It produces state-of-the-art results in speech signal and information processing. In this paper, speech enhancement methods were introduced briefly. Some monaural speech enhancement models based on the deep neural networks were presented. Finally, some views on the development were introduced.

关 键 词：深度学习 语音增强 神经网络 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]