Speech Tokenizer is Key to Consistent Representation
This work addresses the need for consistent speech representation in AI-driven speech processing, offering a versatile tool that improves performance in multiple downstream tasks.
The paper tackled the problem of speech tokenization by introducing a novel tokenizer that simultaneously encodes linguistic and acoustic information, resulting in enhanced representation fidelity across diverse applications like speech coding, voice conversion, and emotion recognition without additional training.
Speech tokenization is crucial in digital speech processing, converting continuous speech signals into discrete units for various computational tasks. This paper introduces a novel speech tokenizer with broad applicability across downstream tasks. While recent advances in residual vector quantization (RVQ) have incorporated semantic elements, they often neglect critical acoustic features. We propose an advanced approach that simultaneously encodes both linguistic and acoustic information, preserving prosodic and emotional content. Our method significantly enhances speech representation fidelity across diverse applications. Empirical evaluations demonstrate its effectiveness in speech coding, voice conversion, emotion recognition, and multimodal language modeling, without requiring additional training. This versatility underscores its potential as a key tool for advancing AI-driven speech processing.