CSTR NAM TIMIT Plus corpus
收藏arXiv2025-09-30 收录
下载链接:
https://nam2speech.github.io/NAM2Speech/
下载链接
链接失效反馈官方服务:
资源简介:
该数据集是一个公开可获取的资源,包含了NAM振动数据以及相应的耳语音频和文本信息。这些资料由一位女性发言者表达了421个句子。该数据集被用于评估语音合成可理解性,其中13%的数据被划为测试用途,5%的数据则用于计算验证损失。整个数据集的规模为421个句子,其任务是对NAM到语音的转换。
This is a publicly available dataset containing NAM vibration data, corresponding whispered audio recordings and text transcripts. It comprises 421 sentences uttered by a female speaker, and is used to evaluate the comprehensibility of speech synthesis. Specifically, 13% of the dataset is allocated for testing purposes, while 5% is utilized for computing validation loss. The core task of this dataset is NAM-to-speech conversion.
提供机构:
CSTR



