使用监督式与自监督语音基础模型对儿童自动语音识别进行基准测试
Benchmarking Children's ASR with Supervised and Self-supervised Speech Foundation Models
AI总结：
该研究针对儿童 ASR 缺乏系统基准的问题，评估多种监督与自监督语音基础模型，并比较数据增强和 PEFT 微调策略，提出 PIF 损失以稳定增强数据微调。
摘要：
语音基础模型（SFM）在监督式系统（例如 Whisper）或自监督系统（例如 WavLM）中，已经在多种语音任务上取得了最先进的结果。然而，SFM 在儿童自动语音识别（ASR）方面的性能尚未得到系统研究。此外，目前还没有采用标准化评估的儿童 ASR 基准，这使得新想法之间的比较十分困难。本文中，我们基于多种 SFM（Whisper、Wav2vec2.0、HuBERT 和 WavLM），在若干儿童语音数据库上发起并提出了一个综合基准。此外，我们通过比较多种数据增强方法和参数高效微调（PEFT）方法，研究了微调策略。我们观察到，当模型规模增大时，这些方法的行为会有所不同。例如，PEFT 在大型模型上能够匹配全量微调的性能，但在小型模型上表现较差。为了稳定使用增强数据进行的微调，我们提出了一种扰动不变微调（PIF）损失作为正则化方法。
英文摘要：
Speech foundation models (SFMs) have achieved state-of-the-art results for various speech tasks in supervised (e.g. Whisper) or self-supervised systems (e.g. WavLM). However, the performance of SFMs for child ASR has not been systematically studied. In addition, there is no benchmark for child ASR with standard evaluations, making the comparisons of novel ideas difficult. In this paper, we initiate and present a comprehensive benchmark on several child speech databases based on various SFMs (Whisper, Wav2vec2.0, HuBERT, and WavLM). Moreover, we investigate finetuning strategies by comparing various data augmentation and parameter-efficient finetuning (PEFT) methods. We observe that the behaviors of these methods are different when the model size increases. For example, PEFT matches the performance of full finetuning for large models but worse for small models. To stabilize finetuning using augmented data, we propose a perturbation invariant finetuning (PIF) loss as a regularization.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
