深度学习语言增强

作者：YeBobr
链接：https://www.zhihu.com/question/273665262/answer/388296862
来源：知乎
著作权归作者所有。商业转载请联系作者获得授权，非商业转载请注明出处。

最近在深度学习在语音增强中的应用最前沿的应该数GAN网络了吧，把生成器当做增强网络，用判别器区分干净语音和增强语音。主要有如下两篇论文：

1.SEGAN: Speech Enhancement Generative Adversarial Network

2.Conditional Generative Adversarial Networks for Speech Enhancement and Noise-Robust Speaker Verification

在卷积神经网络方面，有基于全卷积的，有基于冗余卷积的，在时域上和在频域上处理语音。论文链接如下：

1.Single channel speech enhancement using convolutional neural network

2.A FULLY CONVOLUTIONAL NEURAL NETWORK FOR SPEECH ENHANCEMENT

3.Raw Waveform-based Speech Enhancement by Fully Convolutional Networks

在DNN方面，主要是在频域内处理语音，通过短时傅里叶变换求得短时频谱，然后对短时频谱进行处理，利用含噪语音的相位进行重构增强语音。还有一些小是DNN和传统语音增强方法进行结合的办法，把传统语音中的features换成DNN网络，基本这个套路。论链接如下：

1.Speech Enhancement In Multiple-Noise Conditions using Deep Neural Networks

2.NMF-based Speech Enhancement Incorporating Deep Neural Network

3.A Novel Single Channel Speech Enhancement Based on Joint Deep Neural Network and Wiener Filter

4.An Experimental Study on Speech Enhancement Based on Deep Neural Networks

5.A Regression Approach to Speech Enhancement Based on Deep Neural Networks