数据堂—50小时美国儿童麦克风采集语音数据

该数据由219名美国本地儿童参与录制。录音内容符合儿童特点，主要为故事书、儿歌、口语等内容，每人350句，平均句长4.5次；句子平均重复次数2.1次。采用高保真Blueyeti麦克风录制。文本经过人工转写，准确率高。

This dataset was recorded with 219 local children in the United States. The recording content is tailored to children's characteristics, mainly including storybook readings, nursery rhymes, and spontaneous spoken content. Each participant provided 350 utterances, with an average sentence length of 4.5, and each utterance was repeated an average of 2.1 times. Recordings were made using a high-fidelity Blue Yeti microphone. The corresponding text was manually transcribed with high accuracy.

搜集汇总

背景与挑战

背景概述

该数据集包含50小时美国儿童麦克风采集的语音数据，专为测试儿童英语语音识别模型设计。由219名儿童参与录制，内容涉及故事书、儿歌和口语等，录音格式为高质量WAV文件，并已进行手动转录。