An investigation of the reconstruction capacity of stacked convolutional autoencoders for log-mel-spectrograms

Open in new window