RepCodec: A Speech Representation Codec for Speech Tokenization
This work addresses a bottleneck in speech processing for large language models, offering a robust method for semantic tokenization that enhances information retention, though it is incremental as it builds on existing speech encoders.
The paper tackles the problem of information loss in discrete speech tokenization for large language models by introducing RepCodec, a speech representation codec that learns a vector quantization codebook to reconstruct speech representations, resulting in significant performance improvements over k-means clustering in speech understanding and generation across various encoders and languages.
With recent rapid growth of large language models (LLMs), discrete speech tokenization has played an important role for injecting speech into LLMs. However, this discretization gives rise to a loss of information, consequently impairing overall performance. To improve the performance of these discrete speech tokens, we present RepCodec, a novel speech representation codec for semantic speech tokenization. In contrast to audio codecs which reconstruct the raw audio, RepCodec learns a vector quantization codebook through reconstructing speech representations from speech encoders like HuBERT or data2vec. Together, the speech encoder, the codec encoder and the vector quantization codebook form a pipeline for converting speech waveforms into semantic tokens. The extensive experiments illustrate that RepCodec, by virtue of its enhanced information retention capacity, significantly outperforms the widely used k-means clustering approach in both speech understanding and generation. Furthermore, this superiority extends across various speech encoders and languages, affirming the robustness of RepCodec. We believe our method can facilitate large language modeling research on speech processing.