Neural network for speech removal, which is trained with deep loss of features

摘要

Speech noise removal techniques are provided using a noise removal neural network (NN) that is trained with deep feature losses obtained from an audio classifier NN. A methodology that implements the techniques according to one embodiment includes applying the speech de-noise-NN to be trained to a noisy sample of a training speech signal to generate a processed training speech signal. The method further includes applying a trained audio classifier NN to the processed training speech signal to generate a first set of activation features and applying the trained audio classifier NN to a clear sample of the training speech signal to generate a second set of activation features. The method further includes calculating a loss value based on the first and second set of activation features and performing a back propagation training update of the denoising NN based on the loss value. The process involves iterating this process in order to additionally train the denoising NN.