Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Park, Joonyong, Takamichi, Shinnosuke, Chan, David M., Kando, Shunsuke, Saito, Yuki, Saruwatari, Hiroshi
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2509.01390
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866918133671919616
author Park, Joonyong
Takamichi, Shinnosuke
Chan, David M.
Kando, Shunsuke
Saito, Yuki
Saruwatari, Hiroshi
author_facet Park, Joonyong
Takamichi, Shinnosuke
Chan, David M.
Kando, Shunsuke
Saito, Yuki
Saruwatari, Hiroshi
contents This study presents a comparative analysis of the statistical and linguistic properties of neural audio codecs (NACs). We investigate discrete speech tokens produced by various NAC models, examining their adherence to linguistic statistical laws such as Zipf's law and Heaps' law, as well as their entropy and redundancy. To assess how these token-level properties relate to semantic and acoustic preservation in synthesized speech, we evaluate intelligibility using error rates of automatic speech recognition, and quality using the UTMOS score. Our results reveal that NAC tokens, particularly 3-grams, exhibit language-like statistical patterns. Moreover, these properties, together with measures of information content, are found to correlate with improved performances in speech recognition and resynthesis tasks. These findings offer insights into the structure of NAC token sequences and inform the design of more effective generative speech models.
format Preprint
id arxiv_https___arxiv_org_abs_2509_01390
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Analysing the Language of Neural Audio Codecs
Park, Joonyong
Takamichi, Shinnosuke
Chan, David M.
Kando, Shunsuke
Saito, Yuki
Saruwatari, Hiroshi
Computation and Language
Audio and Speech Processing
This study presents a comparative analysis of the statistical and linguistic properties of neural audio codecs (NACs). We investigate discrete speech tokens produced by various NAC models, examining their adherence to linguistic statistical laws such as Zipf's law and Heaps' law, as well as their entropy and redundancy. To assess how these token-level properties relate to semantic and acoustic preservation in synthesized speech, we evaluate intelligibility using error rates of automatic speech recognition, and quality using the UTMOS score. Our results reveal that NAC tokens, particularly 3-grams, exhibit language-like statistical patterns. Moreover, these properties, together with measures of information content, are found to correlate with improved performances in speech recognition and resynthesis tasks. These findings offer insights into the structure of NAC token sequences and inform the design of more effective generative speech models.
title Analysing the Language of Neural Audio Codecs
topic Computation and Language
Audio and Speech Processing
url https://arxiv.org/abs/2509.01390