Skip to content
CNN that distinguishes between languages
Python Shell
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
__pycache__
grad_CAMS
.gitignore
LICENSE
LangNet_class_index.json
README.md
convert_flac2wav.sh
convert_samplerate.sh
copy_flacFiles.py
create_a_spectrogram_16k.py
create_imgtxt.py
create_imgtxt_16k.py
create_spectrograms.py
create_spectrograms_16k.py
create_test_set.py
create_test_spectrograms.py
filterVis_LangNet.py
gradCAM_LangNet.py
img_set.txt
img_set_16k_train.txt
img_set_16k_val.txt
model.png
model_keras.py
model_keras_16k.py
my_spectrogram.py
remove_spectrograms.sh
stitched_filters_5x5.png
test_set_16k.txt
weights.best.hdf5

README.md

LangNet_CNN

Hrayr Harutyunyan showed that CNNs are very good at identifying what language is being spoken give multiple languages.

This repository is for my exploration of CNNs in language recognition. I am currently using file from several Shtooka databases spanning eight languages. I am still searching for other freely available spoke word databases that contain recordings across many languages and multiple speakers within each language.

The purpose of this work is to examine the features that the CNN architecture deems as important for distinguishing different languages.

During the preprocessing, there are two other folders. One folder containing the Flac files from Shtooka and one folder containing the converted flac - wav files. I have not uploaded these as they each contain over 6000 files.

You can’t perform that action at this time.