DIARISATION👶🏿👶🏾

This repository contains tools that enable diarisation for child language environment analysis.

Dependancies

These tools have been developed and tested in an Ubuntu 20 enviroment using Python 3.8.

 pip install -r requirements.txt

Running speech extractor

 python run.py -s CHI -w False

The argument -s determines the target speaker. The available choices are MOT for mother and CHI for child. The argument -w determines whether to save the spectrogram for the extracted speech utterance or not. Without saving the spectrogram:

 python run.py -s CHI

To save the spectogram:

 python run.py -s CHI -w True

Repository structure

dataset

• Contains a .cha transcript and the corresponding audio file for demonstration.

output

• Holds the results when running experiments with scripts. The extracted utterances and spectrograms are stored under this directory in seperate folders.

run.py

• Script to extract speech utterances for a given speaker according to the timestamps provided on the .cha.

spectrogrammer.py

• Helper script for producing spectograms.

utils.py

• Helper script with various utility functions.