Fetching the paper…

MMDenseLSTM: An efficient combination of convolutional and recurrent neural networks for audio source separation · Around