Towards a Perceptual Loss: Using a Neural Network Codec Approximation as a Loss for Generative Audio Models


Generative audio models based on neural networks have led to considerable improvements across fields including speech enhancement, source separation, and text-to-speech synthesis. These systems are typically trained in a supervised fashion using simple element-wise l1 or l2 losses. However, because they do not capture properties of the human auditory system, such losses encourage modelling perceptually meaningless aspects of the output, wasting capacity and limiting performance. Additionally, while adversarial models have been employed to encourage outputs that are statistically indistinguishable from ground truth and have resulted in improvements in this regard, such losses do not need to explicitly model perception as their task; furthermore, training adversarial networks remains an unstable and slow process. In this work, we investigate an idea fundamentally rooted in psychoacoustics. We train a neural network to emulate an MP3 codec as a differentiable function. Feeding the output of a generative model through this MP3 function, we remove signal components that are perceptually irrelevant before computing a loss. To further stabilize gradient propagation, we employ intermediate layer outputs to define our loss, as found useful in image domain methods. Our experiments using an autoencoding task show an improvement over standard losses in listening tests, indicating the potential of psychoacoustically motivated models for audio generation.


November 2023 | ACM TORS

Unbiased Identification of Broadly Appealing Content Using a Pure Exploration Infinitely-Armed Bandit Strategy

Maryam Aziz, Jesse Anderton, Kevin Jamieson, Alice Wang, Hugues Bouchard, Javed Aslam

October 2023 | CIKM

Graph Learning for Exploratory Query Suggestions in an Instant Search System

Enrico Palumbo, Andreas Damianou, Alice Wang, Alva Liu, Ghazal Fazelnia, Francesco Fabbri, Rui Ferreira, Fabrizio Silvestri, Hugues Bouchard, Claudia Hauff, Mounia Lalmas, Ben Carterette, Praveen Chandar, David Nyhan

September 2023 | RecSys

Accelerating Creator Audience Building through Centralized Exploration

Buket Baran, Guilherme Dinis Junior, Antonina Danylenko, Olayinka S. Folorunso, Gösta Forsum, Maksym Lefarov, Lucas Maystre, Yu Zhao