RF-GML: Reference-Free Generative Machine Listener

TL;DR

RF-GML is a new reference-free audio quality metric that accurately predicts subjective scores for various audio formats and codecs, leveraging transfer learning from a full-reference model.

Contribution

It introduces RF-GML, a novel transfer learning-based reference-free audio quality assessment model with high accuracy and versatility across diverse audio content and codecs.

Findings

Outperforms existing RF models in predicting subjective quality.

Effectively distinguishes different levels of coding artifacts.

Works well across mono, stereo, and binaural audio at 48 kHz.

Abstract

This paper introduces a novel reference-free (RF) audio quality metric called the RF-Generative Machine Listener (RF-GML), designed to evaluate coded mono, stereo, and binaural audio at a 48 kHz sample rate. RF-GML leverages transfer learning from a state-of-the-art full-reference (FR) Generative Machine Listener (GML) with minimal architectural modifications. The term "generative" refers to the model's ability to generate an arbitrary number of simulated listening scores. Unlike existing RF models, RF-GML accurately predicts subjective quality scores across diverse content types and codecs. Extensive evaluations demonstrate its superiority in rating unencoded audio and distinguishing different levels of coding artifacts. RF-GML's performance and versatility make it a valuable tool for coded audio quality assessment and monitoring in various applications, all without the need for a reference signal.