consynthance

Studying consonance as a result of vocal similarity

<img src="img/en100.png" width="500"/> Mean spectrum at different frequency ratios from 100 examples of english speech.

Usage

You'll need to install essentia, which has a platform dependant installation process.
(See here.)

On macOS you can do this with homebrew.

brew tap MTG/essentia
brew install essentia --HEAD --without-python@2

If you are using virtualenv make sure to include site packages for this to work.

python3 -m venv env/ --system-site-packages 

Install requirements.

pip install -r requirements.txt

Launch the notebooks.

jupyter notebook

Note: developed with Python 3.7 and requires Python 3.6 or above.

Data

To start we are using the Common Voice dataset from Mozilla, which provides nearly 70GB of speech data from 30 different languages.

Downloading this data is easy with the provided bash script. Run the data.sh bash script as follows, where the command line argument specifies a directory to store the data. This will take a while to complete, probably in the range of 2-4 hrs depending on your downloaded speed.

$ ./data.sh /dir/to/store/output