HSD
收藏资源简介:
HSD数据集由早稻田大学信息生产与系统研究生院创建,包含68首来自YouTube的流行歌曲,旨在提供音乐的层次结构信息。该数据集详细记录了每首歌曲中每个音乐音符的起始/结束时间、音高、持续时间和歌词,采用增强的LyRiCs格式。创建过程分为两个阶段:首先使用音乐符号和LyRiCs文件创建初始标签,然后通过手动校准这些标签以确保准确性。HSD数据集主要应用于音乐信息检索领域,特别是自动歌唱转录任务,以提高转录的准确性和效率。
The HSD Dataset was developed by the Graduate School of Information, Production and Systems, Waseda University. It contains 68 popular songs sourced from YouTube, and is intended to provide hierarchical structural information of music. This dataset meticulously records the start/end time, pitch, duration, and lyrics of each musical note across every song, utilizing an enhanced LyRiCs format. Its creation process is divided into two stages: firstly, initial labels are generated using musical notation and LyRiCs files, followed by manual calibration of these labels to ensure accuracy. The HSD Dataset is primarily applied in the field of Music Information Retrieval (MIR), specifically for automatic singing transcription tasks to enhance the accuracy and efficiency of transcription.

- 1HSD: A hierarchical singing annotation dataset早稻田大学信息生产与系统研究生院 · 2022年



