Amateur Vocal Percussion (AVP) dataset

AVP数据集是由罗利有限公司和伦敦玛丽女王大学合作创建，专注于无或少量经验的业余声音打击分析。该数据集包含9780条由28名参与者录制的语音模仿打击乐的音频，每条音频均标注了起始时间和标签（如鼓、军鼓、闭合和高开合的踩镲）。数据集的创建旨在帮助算法更好地泛化不同参与者，提高性能。AVP数据集的应用领域主要集中在音乐信息检索，特别是声音打击乐的自动映射，以加速节奏原型的创建。

The AVP Dataset was developed in collaboration between Raleigh Limited and Queen Mary University of London, focusing on amateur percussive sound analysis for people with little or no prior relevant experience. This dataset contains 9,780 audio recordings of vocal imitations of percussion instruments made by 28 participants, with each audio clip annotated with its start time and corresponding labels such as drum, snare drum, closed hi-hat, and open hi-hat. The dataset was created to help algorithms better generalize across different participants and improve their performance. The main application areas of the AVP Dataset are music information retrieval, particularly automatic mapping of percussive sounds, to accelerate the creation of rhythm prototypes.