PSCodec: A Series of High-Fidelity Low-bitrate Neural Speech Codecs Leveraging Prompt Encoders

TL;DR

This paper introduces PSCodec, a series of neural speech codecs using prompt encoders that achieve high-quality speech reconstruction at low bitrates, advancing speech compression technology.

Contribution

The paper presents three novel neural speech codecs leveraging prompt encoders, including a new disentanglement method and an attention network to improve low-bitrate speech quality.

Findings

All three codecs outperform state-of-the-art neural codecs in quality and speaker similarity.

PSCodec-DRL-ICT achieves high performance but requires extensive tuning.

PSCodec-CasAN offers a less labor-intensive alternative with comparable results.

Abstract

Neural speech codecs have recently emerged as a focal point in the fields of speech compression and generation. Despite this progress, achieving high-quality speech reconstruction under low-bitrate scenarios remains a significant challenge. In this paper, we propose PSCodec, a series of neural speech codecs based on prompt encoders, comprising PSCodec-Base, PSCodec-DRL-ICT, and PSCodec-CasAN, which are capable of delivering high-performance speech reconstruction with low bandwidths. Specifically, we first introduce PSCodec-Base, which leverages a pretrained speaker verification model-based prompt encoder (VPP-Enc) and a learnable Mel-spectrogram-based prompt encoder (MelP-Enc) to effectively disentangle and integrate voiceprint and Mel-related features in utterances. To further enhance feature utilization efficiency, we propose PSCodec-DRL-ICT, incorporating a structural similarity (SSIM) based disentangled representation loss (DRL) and an incremental continuous training (ICT) strategy. While PSCodec-DRL-ICT demonstrates impressive performance, its reliance on extensive hyperparameter tuning and multi-stage training makes it somewhat labor-intensive. To circumvent these limitations, we propose PSCodec-CasAN, utilizing an advanced cascaded attention network (CasAN) to enhance representational capacity of the entire system. Extensive experiments show that our proposed PSCodec-Base, PSCodec-DRL-ICT, and PSCodec-CasAN all significantly outperform several state-of-the-art neural codecs, exhibiting substantial improvements in both speech reconstruction quality and speaker similarity under low-bitrate conditions.