Speech Separation using Neural Audio Codecs with Embedding Loss

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Yip, Jia Qi, Kwok, Chin Yuen, Ma, Bin, Chng, Eng Siong
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866929606260424704
author Yip, Jia Qi
Kwok, Chin Yuen
Ma, Bin
Chng, Eng Siong
author_facet Yip, Jia Qi
Kwok, Chin Yuen
Ma, Bin
Chng, Eng Siong
contents Neural audio codecs have revolutionized audio processing by enabling speech tasks to be performed on highly compressed representations. Recent work has shown that speech separation can be achieved within these compressed domains, offering faster training and reduced inference costs. However, current approaches still rely on waveform-based loss functions, necessitating unnecessary decoding steps during training. We propose a novel embedding loss for neural audio codec-based speech separation that operates directly on compressed audio representations, eliminating the need for decoding during training. To validate our approach, we conduct comprehensive evaluations using both objective metrics and perceptual assessment techniques, including intrusive and non-intrusive methods. Our results demonstrate that embedding loss can be used to train codec-based speech separation models with a 2x improvement in training speed and computational cost while achieving better DNSMOS and STOI performance on the WSJ0-2mix dataset across 3 different pre-trained codecs.
format Preprint
id arxiv_https___arxiv_org_abs_2411_17998
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Speech Separation using Neural Audio Codecs with Embedding Loss
Yip, Jia Qi
Kwok, Chin Yuen
Ma, Bin
Chng, Eng Siong
Audio and Speech Processing
Neural audio codecs have revolutionized audio processing by enabling speech tasks to be performed on highly compressed representations. Recent work has shown that speech separation can be achieved within these compressed domains, offering faster training and reduced inference costs. However, current approaches still rely on waveform-based loss functions, necessitating unnecessary decoding steps during training. We propose a novel embedding loss for neural audio codec-based speech separation that operates directly on compressed audio representations, eliminating the need for decoding during training. To validate our approach, we conduct comprehensive evaluations using both objective metrics and perceptual assessment techniques, including intrusive and non-intrusive methods. Our results demonstrate that embedding loss can be used to train codec-based speech separation models with a 2x improvement in training speed and computational cost while achieving better DNSMOS and STOI performance on the WSJ0-2mix dataset across 3 different pre-trained codecs.
title Speech Separation using Neural Audio Codecs with Embedding Loss
topic Audio and Speech Processing
url https://arxiv.org/abs/2411.17998