An Expert Ground Truth Set for Audio Chord Recognition and Music Analysis
An Expert Ground Truth Set for Audio Chord Recognition and Music Analysis
复制标题
用于音频和弦识别和音乐分析的专家地面实况集
DOI：
--
复制
发表时间：
2011
期刊：
影响因子：
--
通讯作者：
Ichiro Fujinaga
中科院分区：
文献类型：
--
作者：
J. Burgoyne;Jonathan Wild;Ichiro Fujinaga
近年来，音频和弦识别引起了人们的极大兴趣，但在数量和采样范围方面严重缺乏可靠的训练数据阻碍了进展。我们与一支训练有素的爵士音乐家团队合作，从1958年至1991年美国公告牌“Hot 100”排行榜上随机选出的一千多首歌曲中收集了和声的时间对齐transanterior。这些音律包含了关于上部扩展和变化的完整信息，以及关于节拍、乐句和更大的音乐结构的信息。我们预计，这些transmittance将使音频和弦识别算法的训练质量的显着进步，此外，由于创新的采样方法，数据是可用的，因为它们代表计算音乐学。本文包括一些摘要数字和统计数据，以帮助读者了解数据的范围，以及为自己的研究获得数据库的信息。
Audio chord recognition has attracted much interest in recent years, but a severe lack of reliable training data—both in terms of quantity and range of sampling—has hindered progress. Working with a team of trained jazz musicians, we have collected time-aligned transcriptions of the harmony in more than a thousand songs selected randomly from the Billboard “Hot 100” chart in the United States between 1958 and 1991. These transcriptions contain complete information about upper extensions and alterations as well as information about meter, phrase, and larger musical structure. We expect that these transcriptions will enable significant advances in the quality of training for audio-chord-recognition algorithms, and furthermore, because of an innovative sampling methodology, the data are usable as they stand for computational musicology. The paper includes some summary figures and statistics to help readers understand the scope of the data as well as information for obtaining the transcriptions for their own research.