Experiment dataset supplementary materials for DLfM 2018 submission

This is the experiment dataset supplementary materials for the DLfM 2018 paper submission:

An extended jingju solo singing voice dataset and its application on automatic assessment of singing pronunciation and overall quality at phoneme-level

Files: 

1.  dlfm_experiment_dataset_file_list.ods: recording file names of train, validation and test sets split.
2.  dlfm_experiment_dataset_phoneme_numbers.ods: phoneme numbers of each phone class in train, validation and test sets.
3.  freesound_extractor_feature_list.ods: freesoundExtractor feature name list used in ANOVA feature analysis.
4.  log-mel-scaler-keys-label-encoder.zip: files required for training the embedding model, includes logarithmic features, feature scaler, phoneme dictionary keys and label encoder.
5. anova_analysis_essentia_feature.zip: Essentia freesoundExtractor features of each phoneme for ANOVA analysis.

The recordings listed in dlfm_experiment_dataset_file_list.ods are taken from a collection of jingju solo singing voice audio datasets, which contains three parts:

• Part 1: https://doi.org/10.5281/zenodo.780559
• Part 2: https://doi.org/10.5281/zenodo.842229
• Part 3: https://doi.org/10.5281/zenodo.1244732

Contact information:

If you have any question, please contact the authors:

Rong Gong: Email - rong<dot>gong<at>upf<dot>edu