MUSERC Pre-Processing

Pre-process the raw MUSERC dataset so that the audio, video and sensor data is cut, cleaned up and compressed for easier and faster handling.

Audio

• Recorded using AKG C414
• Sampling Rate 48000 Hz
• Raw Data: 24bit uncompressed mono PCM
• Pre-Processed Data: 16bit uncompressed mono PCM

Video

• 2000fps high speed camera
• Focusing on the string to extract the movement pattern
• Resolution 1024x768
• Raw Data: H264 lossless mode
• Pre-Processed Data: H264 lossy compression

Sensors

• Measured finger position using a linear membrane potentiometer.
• Recorded using 12bit D/A at 750Hz.

Running the Pre-Processing

Video processing requires ffmpeg with x264.

• Run make

make installs pip requirements and fetches the dataset (~4gb).

Authors

• Fabian-Robert Stöter, International Audio Laboratories Erlangen
• Michael Müller, TU Graz
• Bernd Edler,International Audio Laboratories Erlangen

Licence

MUSERC is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. If you use this dataset in your academic research, please cite our paper:

Fabian-Robert Stöter, Michael Müller, and Bernd Edler. 2015. Multi-Sensor Cello Recordings for Instantaneous Frequency Estimation. In Proceedings of the 23rd Annual ACM Conference on Multimedia Conference (MM '15). ACM, New York, NY, USA, 995-998.