CCOM-HuQin: an Annotated Multimodal Chinese Fiddle Performance Dataset

NOTICE: The access of CCOM-HuQin-v2.0.1, the latest version, is opening.  You may update the v2.0 with the v2.0.1.

HuQin(胡琴) is a family of traditional Chinese fiddles (bowed string instruments). CCOM-HuQin is the first multimodal performance dataset of HuQin music, which encompasses eight representative HuQin categories played by professional players majoring in HuQin performing arts at Central Conservatory of Music (CCOM).

The instruments are as follows:

• Erhu (二胡)
• Banhu family (板胡): Soprano Banhu (高音板胡), Alto Banhu (中音板胡), Tenor Banhu (次中音板胡) and Bass Banhu (低音板胡)
• Gaohu (高胡)
• Zhuihu (坠胡)
• Zhonghu (中胡)

In the dataset, we highlight the rich and complex Playing Techniques(PTs), which constitute important aspects of distinct regional and ethnic features in the HuQin family. CCOM-HuQin comprises two subsets: (1) single PT recordings that include various articulations regarding velocity, dynamics, and pitch intervals; (2) classical musical excerpts provided with performance scores and transcription files with ground-truth note-level onset and frame-level pitch annotations validated by professionals. All annotation files and performance scores are provided in the CCOM-HuQin-v2.0-audios.zip folder. For each musical piece, we provide high-quality audio and aligned videos with multiple-camera views.

Overall, the CCOM-HuQin dataset contains audio-visual recordings of over 12,000 single PT clips and 57 annotated musical excerpts. In the previous version 1.1, we open access to 845 single PT clips and 10 annotated musical excerpts, a subset that covers all the PT categories and instruments. 

CCOM-HuQin dataset is developed by Yu Zhang (PhD candidate) and Ziya Zhou (graduate), under the supervision of Prof. Maosong Sun, Prof. Feng Yu and Prof. Xiaobing Li, at the Department of AI Music and Music Information Technology, CCOM. We thank Kai Chen, Li Li, Zengqi Zhang, Runze Li, Peize Li, Zhaoxing Wang, Chuting Wang for participating in performance and annotation and providing professional advice on building the dataset, Yiru Xu and Xueying Wang for their recording expertise. We also thank Prof. Zijin Li for providing constructive comments. 

For more details about the usage of the dataset, please refer to the readme-v2.0.pdf in the dataset folder CCOM-HuQin-v2.0-audios.zip. Code for data processing is available on GitHub.

This dataset is available under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.

Related paper:

Zhang, Y., Zhou, Z., Li, X., Yu, F., & Sun, M. (2023). CCOM-HuQin: An Annotated Multimodal Chinese Fiddle Performance Dataset. Transactions of the International Society for Music Information Retrieval, 6(1), 60–74. DOI: https://doi.org/10.5334/tismir.146

Contact

yuzhang93@126.com

dange_zzy@outlook.com