This code accompanies the paper "Spectral Visibility Graphs: Application to Similarity of Harmonic Signals" by Delia Fano Yela, Dan Stowell and Mark Sandler (EUSIPCO 2019) where we introduce the visibility graph for audio spectra and propose a novel representation for audio analysis: the spectral visibility graph degree.Such representation inherently captures the harmonic content of the signal whilst being resilient to broadband noise. We present experiments demonstrating its utility to measure robust similarity between harmonic signals in real and synthesised audio data.

In the paper we present two experiments demonstrating the utility of the proposed representation of audio signals for harmonic similarity measure.

Experiment 01 for synthesised audio data:

• Audio data used: AUDIO/synth_dataset
• Script to run experiment: experiment01.py
• Script to plot the results (Figure 3 in the paper): plot_exp01.py

Experiment 02 for real audio data:

• Audio data used: DSD100 dataset available at https://sigsep.github.io/datasets/dsd100.html
• Script to run experiment: experiment02.py
• Script to plot the results (Figure 4 in the paper): plot_exp02.py

Other:

• plot_representation.py : Script to plot Figure 2 of the paper showing an example of a spectrogram and its corresponding spectral visibility graph degree proposed representation. The sample audio can be found in AUDIO/sample_used_plot_representation

• visibility_algorithms.py : our implementations of the different visibility graphs algorithms.

• figures : folder containing the images used in the paper

• results_experiments : folder containing the results (in form .csv) obtained from the experiments scripts and used by the plotting scripts.

If you use this work for your research please cite:

@INPROCEEDINGS{vgspectra,  
  author={D. F. {Yela} and D. {Stowell} and M. {Sandler}},  
  booktitle={2019 27th European Signal Processing Conference (EUSIPCO)},  
  title={Spectral Visibility Graphs: Application to Similarity of Harmonic Signals},   
  year={2019},  
  volume={},  
  number={},
  pages={1-5}}

AUTHOR: Delia Fano Yela
DATE: April 2020 CONTACT: [email protected] - most recently : [email protected]