BIAOBEI

该数据集名为BIAOBEI，包含了10,000句普通话语音句子，每句都附有相应的音频记录和文本注释。整个数据集的有效语音时长约为12小时。数据集被划分为训练集（9,400句）、验证集（500句）和测试集（100句）。音频的采样率为22.05kHz，特征提取采用12.5毫秒的帧移大小和50毫秒的窗口大小。该数据集的规模为10,000句，其任务是语音合成。

This dataset, named BIAOBEI, consists of 10,000 Mandarin speech sentences, each paired with its corresponding audio recording and textual annotation. The total effective speech duration of the entire dataset is approximately 12 hours. It is divided into three subsets: the training set with 9,400 sentences, the validation set with 500 sentences, and the test set with 100 sentences. The audio sampling rate is 22.05 kHz, and a frame shift size of 12.5 ms and a window size of 50 ms are adopted for feature extraction. With a scale of 10,000 sentences, the target task of this dataset is speech synthesis.

搜集汇总

背景与挑战

背景概述

该数据集是标贝科技提供的中文标准女声音库，包含约12小时的有效时长，录音环境专业，语料涵盖多个领域，适用于科研、生活出行、智能科技等场景。数据集提供了详细的标注内容和技术参数，支持非商业用途。