To speed up the optimization process, we transform the corresponding problem into a lower-dimensional latent space learned by a variational autoencoder. This is trained on a total of 6839 different 2D ...

This article focuses on leveraging deep representation learning (DRL) for speech enhancement (SE). In general, the performance of the deep neural network (DNN) is heavily dependent on the learning of ...