Jamendo Corpus

Jamendo Corpus 是一个语音检测数据集，由来自 Jamendo 免费音乐共享网站的具有知识共享许可的 93 首歌曲组成。每首歌曲的片段都被注释为“语音”（唱歌或口语）或“无语音”。这些歌曲总共构成了大约 6 小时的音乐。这些文件都来自不同的艺术家，代表了主流商业音乐的各种流派。 Jamendo 音频文件以 112KB/s 比特率的立体声 Vorbis OGG 44.1kHz 编码。原始拆分分别在训练、验证和测试集中包含 61、16 和 16 首歌曲。The Jamendo Corpus is a speech detection dataset composed of 93 Creative Commons-licensed songs sourced from the Jamendo free music sharing website. Each segment of these songs is annotated as either "speech" (singing or spoken language) or "non-speech". These songs collectively total approximately 6 hours of audio content. All audio files originate from diverse artists and represent various genres of mainstream commercial music. Jamendo audio files are encoded as stereo Vorbis OGG format at a 44.1 kHz sampling rate and 112 KB/s bitrate. The original dataset split allocates 61, 16, and 16 songs to the training, validation, and test sets respectively.

搜集汇总

背景与挑战

背景概述

Jamendo Corpus是一个用于语音检测的数据集，包含来自Jamendo网站的93首知识共享许可歌曲，总时长约6小时，每首歌曲的片段被标注为语音或非语音。数据集以Vorbis OGG格式提供，原始拆分包括训练、验证和测试集，适用于音频识别和语音事件检测任务。