基于混合深度神经网络的语音增强方法研究

Research on speech enhancement method based on hybrid deep neural networks

针对基于混合深度神经网络的语音增强方法展开研究,阐述了该方法提出的背景、模型原理和实施过程。搭建了基于混合深度神经网络的深度学习语音增强模型,并与仅基于DNN的语音增强模型进行了对比实验,验证了基于混合深度神经网络的语音增强方法,进一步提高了增强语音的质量。

The background,model principle and implementation process of speech enhancement based on hybrid deep neural networks are systematically expounded. A hybrid DNN-based deep learning speech enhancement model was built and compared with the stand alone DNN-based model. Experimental results show that the speech enhancement method based on hybrid DNN further improves the quality of enhanced speech.