Google Speech Commands Dataset (V1 and V2)

该数据集分为两个版本：版本1（V1）包含30个简短指令，由1881位发言者录制；版本2（V2）包含35个指令，由2618位发言者录制。数据集涵盖了12类和35类的分类任务，其中包括如“是”、“否”、“上”、“下”、“左”、“右”、“开”、“关”、“停止”、“前进”等指令，以及静默和未知类别。在规模上，GSC-V1拥有56588个训练样本、7743个验证样本和7835个测试样本；而GSC-V2则包含92843个训练样本、11003个验证样本和12005个测试样本。该数据集的任务是关键词识别（Kws）。

This dataset has two versions: Version 1 (V1) contains 30 short commands recorded by 1881 speakers; Version 2 (V2) contains 35 commands recorded by 2618 speakers. The dataset covers 12-class and 35-class classification tasks, including commands such as 'Yes', 'No', 'Up', 'Down', 'Left', 'Right', 'On', 'Off', 'Stop', 'Forward', as well as silence and unknown classes. In terms of scale, GSC-V1 has 56588 training samples, 7743 validation samples, and 7835 test samples; while GSC-V2 contains 92843 training samples, 11003 validation samples, and 12005 test samples. The task of this dataset is keyword spotting (KWS).