Beijing Opera Percussion Instrument Dataset

The Beijing Opera percussion instrument dataset is a collection of audio examples of individual strokes spanning the four percussion instrument classes used in Beijing Opera (Jingju, 京剧).

Beijing Opera uses six main percussion instruments that can be grouped into four classes: 

  1. Bangu (Clapper-drum) consisting of Ban (the clapper, a wooden board-­shaped instrument) + danpigu (a wooden drum struck by two wooden sticks)
  2. Naobo (Cymbals) consisting of two cymbal instruments Qibo+Danao
  3. Daluo: Large gong
  4. Xiaoluo: Small gong


This dataset can be downloaded here.

Each audio file is named as, 


Audio content

The dataset provides audio examples for each of these instrument classes. The number of examples is shown in Table 1. 
1. The dataset
Instrument class Bangu Daluo Naobo Xiaoluo Total
No. of examples 59 50 62 65 236
The audio examples were recorded under studio conditions by Mi Tian at the Centre for Digital Music, Queen Mary University of London, UK in September 2013 using an AKG C414 microphone. The audio was sampled at 44.1 kHz and stored as 16 bit wav files. The instruments were played by Ying Wan of the London Jing Kun Opera Association. Unlike some instruments that can be tuned, these percussion instruments are made from metal casting. Thus, there can be subtle timbral differences even across different instruments of the same kind. For each of these instruments, we used 2-3 individual instruments to record the samples, hoping to achieve a better timbre coverage. Further, audio samples were recorded using different playing techniques for each instrument. 
The dataset can be used for training models for each percussion instrument class. 


The dataset was used as the training dataset in the following paper. Please cite it if you use the dataset in your work:

Mi Tian, Ajay Srinivasamurthy, Mark Sandler, and Xavier Serra, "A Study of Instrument-wise Onset Detection in Beijing Opera Percussion Ensembles", in Proceedings of ICASSP 2014, Florence, Italy, May 2014. (to appear)


If you have any questions or comments about the dataset, please feel free to write to us: 

Mi Tian () or Ajay Srinivasamurthy ()