rpetit/audio-event-detection

Detection of events described by hidden (semi-)Markov models in an audio stream

Audio event detection

Work conducted during an internship at IRCAM.

Disclaimer

This code is no longer actively maintained. The algorithms implemented in this library were only tested on toy examples and did not yield satisfying performances on most experiments. Audio files in the examples/data folder seem corrupted, and example scripts can therefore no longer be ran.

pip installation

Run the following command :

pip install -e git+https://github.com/rpetit/audio-event-detection#egg=evdetect

Example

Detection of a bird's song with a hidden Markov model.

From the examples directory, run bird.py to get the following output :

This will also output a .wav file in examples/results where only the segments matching the event model have been preserved.

See this page for more examples.

References

[1] A. Bietti, online learning for audio clustering and segmentation Master thesis, 2014

[2] A. Cont, a coupled duration-focused architecture for realtime music to score alignment, IEEE Transactions on Pattern Analysis and Machine Intelligence, 2010

[3] Y. Matsubara, Y. Sakurai, N. Ueda and M. Yoshikawa, fast and exact monitoring of co-evolving data streams, IEEE International Conference on Data Mining, 2014

[4] L. Rabiner, a tutorial on hidden Markov models and selected applications in speech recognition, Proceedings of the IEEE, 1989