This is the music vibrato dataset with ground truth for vibrato detection and analysis. If you are using this dataset, please cite Luwei Yang, Khalid Z. Rajab and Elaine Chew. Filter Diagonalisation Method for Music Signal Analysis: Frame-wise Vibrato Detection and Estimation, Journal of Mathematics and Music, DOI: 10.1080/17459737.2016.1263897, 2017.
Areas_and_parameters has two instruments, erhu (Chinese two-stringed instrument) and violin. It has ground truth indicating the vibrato areas and also the indications of each half cycle within a vibrato. The latter can be used for calculating vibrato rates and extents. The area ground truth is named as AudioName-Annotation-(new).csv, and the half cycle indication ground truth is named as AudioName-Annotation-Stat.csv.
Areas_only has only the ground truth indicating the vibrato areas. It contains CMMSD and Coler2011 dataset.
-
CMMSD wav files is created by Henrik von Coler, I created the vibrato ground truth. If you use it please cite von Coler, Henrik, and Alexander Lerch. CMMSD: A Data Set for Note-Level Segmentation of Monophonic Music." Audio Engineering Society Conference: 53rd International Conference: Semantic Audio. Audio Engineering Society, 2014.
-
Coler2011 is the vibrato dataset used in Von Coler, Henrik, and Axel Röbel. "Vibrato detection using cross correlation between temporal energy and fundamental frequency." Audio Engineering Society Convention 131. Audio Engineering Society, 2011. The vibrato ground truth is created by von Coler, named as AudioName.txt. I attached a matlab file transferColerTruthData.m to transfer this file into ordinary format of AudioName-Annotation-(new).csv.
Description of some files
- AudioName-Annotation-(new).csv: the area ground truth. You only need to use the first column showing the starting point of each vibrato, and the third column showing the duration.
- AudioName-Annotation-Stat.csv: the half cycle indication ground truth
- AudioName-Yin.csv: the f0 get from pyin
TODO: Erhu/Wangguotong and Violin/Laurel lack of half cycle annotation for vibrato rate and extent estimation.