Speech Commands
收藏资源简介:
语音命令是一个语音单词的音频数据集,旨在帮助训练和评估关键字识别系统。该数据集 (1.4 GB) 65,000了30个短词的一秒钟长话语,由数千个不同的人提供,由公众通过AIY网站提供。这是一套一秒的。wav音频文件,每个文件都包含一个口语单词。这些单词来自一小部分命令,并由各种不同的说话者说出。音频文件会根据它们包含的单词组织到文件夹中,并且此数据集旨在帮助训练简单的机器学习模型。
Speech Commands is an audio dataset of spoken words designed to help train and evaluate keyword spotting systems. With a size of 1.4 GB, this dataset comprises 65,000 one-second long utterances covering 30 short words, contributed by thousands of different speakers via the AIY website. This dataset consists of one-second .wav audio files, each containing a single spoken word. These words belong to a small set of command terms, and were uttered by a diverse group of speakers. The audio files are organized into folders based on the specific word they contain, and this dataset is intended to aid in training simple machine learning models.

- Speech Commands数据集首次发布,包含65,000个简短的语音命令录音,涵盖30个不同的单词。
- Speech Commands数据集在Google AI Blog上正式介绍,并开始被广泛应用于语音识别模型的训练和评估。
- Speech Commands数据集的扩展版本发布,增加了更多的语音样本和新的语言类别,进一步丰富了数据集的内容。
- Speech Commands数据集被多个研究团队用于开发和测试新的语音识别算法,推动了语音技术的发展。
- 1Speech Commands: A Dataset for Limited-Vocabulary Speech RecognitionGoogle · 2018年
- 2Efficient Keyword Spotting Using Dilated Convolutions and GatingUniversity of Oxford · 2019年
- 3Small-Footprint Keyword Spotting Using Deep Neural NetworksUniversity of Waterloo · 2019年
- 4A Comparative Study of Deep Learning Models for Keyword SpottingUniversity of California, Irvine · 2020年
- 5Improving Keyword Spotting through Attention Mechanisms and Data AugmentationUniversity of Michigan · 2021年



