sliding HPSS

Coded by : Hideyuki Tachibana

What's this?

This is an example of the implementation of "sliding HPSS" described in the following paper.

• N. Ono, K. Miyamoto, H. Kameoka, S. Sagayama "A real-time equalizer of harmonic and percussive components in music signals" in Proc. ISMIR 2008.

This program is coded for following paper (IEEE Xplore) on singing voice enhancement, in which it is applied twice with different parameter sets.

@article{tachibana2014singing,
  author={Hideuyki Tachibana and Nobutaka Ono and Shigeki Sagayama},
  journal={IEEE/ACM Transactions on Audio, Speech, and Language Processing}, 
  title={Singing Voice Enhancement in Monaural Music Signals Based on 
  	Two-stage Harmonic/Percussive Sound Separation on Multiple Resolution Spectrograms}, 
  year={2014},
  volume={22},
  number={1},
  pages={228-237}
}

See also the following conferene paper (IEEE Xplore).

@inproceedings{tachibana2010melody,
  author={Hideyuki Tachibana and Tatsuma Ono and Nobutaka Ono and Shigeki Sagayama},
  booktitle={2010 IEEE International Conference on Acoustics, Speech and Signal Processing}, 
  title={Melody line estimation in homophonic music audio signals 
  	based on temporal-variability of melodic source}, 
  year={2010},
  pages={425-428}
}

How to use (on UNIX like system.)

Build

Download the codes as follows.

cd your_working_directory
git clone https://github.com/tachi-hi/slidingHPSS

Install FFTW3 (libfftw3) in your system. If you are an Ubuntu user,

apt-get install libfftw3-dev

Then type the follwoing commands to build the code and execute the obtained binary.

bash setup.sh
./bin/slidingHPSS -h

then usage massage will be displayed.

Preprocessing: Stereo 44.1kHz MP3 -> Mono 16kHz WAV

YourMP3file.mp3 is converted to a suitable format by following commands.

mpg123 -w wavefile_tmp.wav yourMP3file.mp3
sox wavfile_tmp.wav -r 16000 -c 1 input.wav
rm wavfile_tmp.wav

Singing Voice Enhancement

Just run the script file twostageHPSSscripts/twostageHPSS.sh. The file named input.wav in the working directory will be separated into three files named _H.wav, _P.wav, and _V.wav.

Singing Voice Suppression

You can similarly obtain a "karoake" signal just by mixing the two files _H.wav and _P.wav as follows.

sox -m _H.wav _P.wav _karaoke.wav

See also euterpe, an automatic audio-to-audio karaoke system.

License

Free

Academic Use

Please cite one of the papers above.

Links

Other implementations of HPSS

• Our group
  • (non-sliding) HPSS
  • vocal suppression based on multi-stage HPSS ("euterpe")
  • HPSS with GUI (http://hil.t.u-tokyo.ac.jp/software/) ... [expired link]
• Others
  • librosa: This audio library for Python contains a variant of HPSS.

Data

• MIR-1K