Multi-Sampling-Frequency Naturalness MOS Prediction Using Self-Supervised Learning Model with Sampling-Frequency-Independent Layer

TL;DR

This paper presents a self-supervised learning model with an SF-independent layer for predicting speech naturalness across multiple sampling frequencies, achieving top rankings in the AudioMOS Challenge 2025.

Contribution

It introduces an SF-independent convolutional layer within an SSL model for robust MOS prediction across various sampling frequencies, with strategies like knowledge distillation and large-scale pretraining.

Findings

Ranked first in one evaluation metric at AMC 2025

Achieved fourth place in overall ranking

Demonstrated effectiveness of SF-independent features

Abstract

We introduce our submission to the AudioMOS Challenge (AMC) 2025 Track 3: mean opinion score (MOS) prediction for speech with multiple sampling frequencies (SFs). Our submitted model integrates an SF-independent (SFI) convolutional layer into a self-supervised learning (SSL) model to achieve SFI speech feature extraction for MOS prediction. We present some strategies to improve the MOS prediction performance of our model: distilling knowledge from a pretrained non-SFI-SSL model and pretraining with a large-scale MOS dataset. Our submission to the AMC 2025 Track 3 ranked the first in one evaluation metric and the fourth in the final ranking. We also report the results of our ablation study to investigate essential factors of our model.