Saved in:
Bibliographic Details
Main Authors: Dong, Zhongren, Wang, Bin, Han, Jing, Guo, Haotian, Mo, Xiaojun, Cao, Yimin, Zhang, Zixing
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2512.20944
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914218464247808
author Dong, Zhongren
Wang, Bin
Han, Jing
Guo, Haotian
Mo, Xiaojun
Cao, Yimin
Zhang, Zixing
author_facet Dong, Zhongren
Wang, Bin
Han, Jing
Guo, Haotian
Mo, Xiaojun
Cao, Yimin
Zhang, Zixing
contents Neural Speech Codecs face a fundamental trade-off at low bitrates: preserving acoustic fidelity often compromises semantic richness. To address this, we introduce SACodec, a novel codec built upon an asymmetric dual-quantizer that employs our proposed Semantic Anchoring mechanism. This design strategically decouples the quantization of Semantic and Acoustic details. The semantic anchoring is achieved via a lightweight projector that aligns acoustic features with a frozen, large-scale mHuBERT codebook, injecting linguistic priors while guaranteeing full codebook utilization. Sequentially, for acoustic details, a residual activation module with SimVQ enables a single-layer quantizer (acoustic path) to faithfully recover fine-grained information. At just 1.5 kbps, SACodec establishes a new state of the art by excelling in both fidelity and semantics: subjective listening tests confirm that its reconstruction quality is perceptually highly comparable to ground-truth audio, while its tokens demonstrate substantially improved semantic richness in downstream tasks.
format Preprint
id arxiv_https___arxiv_org_abs_2512_20944
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle SACodec: Asymmetric Quantization with Semantic Anchoring for Low-Bitrate High-Fidelity Neural Speech Codecs
Dong, Zhongren
Wang, Bin
Han, Jing
Guo, Haotian
Mo, Xiaojun
Cao, Yimin
Zhang, Zixing
Sound
Neural Speech Codecs face a fundamental trade-off at low bitrates: preserving acoustic fidelity often compromises semantic richness. To address this, we introduce SACodec, a novel codec built upon an asymmetric dual-quantizer that employs our proposed Semantic Anchoring mechanism. This design strategically decouples the quantization of Semantic and Acoustic details. The semantic anchoring is achieved via a lightweight projector that aligns acoustic features with a frozen, large-scale mHuBERT codebook, injecting linguistic priors while guaranteeing full codebook utilization. Sequentially, for acoustic details, a residual activation module with SimVQ enables a single-layer quantizer (acoustic path) to faithfully recover fine-grained information. At just 1.5 kbps, SACodec establishes a new state of the art by excelling in both fidelity and semantics: subjective listening tests confirm that its reconstruction quality is perceptually highly comparable to ground-truth audio, while its tokens demonstrate substantially improved semantic richness in downstream tasks.
title SACodec: Asymmetric Quantization with Semantic Anchoring for Low-Bitrate High-Fidelity Neural Speech Codecs
topic Sound
url https://arxiv.org/abs/2512.20944