Overview / Usage

A deep learning will be trained for selected languages and also classified into geographic locations. To build the deep learning model, various techniques such as filtering, compression, feature extraction will be considered. Also, DeepSpeech machine learning techniques will be relevant for the implementation. Speech data will be sourced from known native speakers from selected regions.

Methodology / Approach

Mel frequency cepstral coefficients (MFCCs) will be used for audio extraction. Neural Networks will be useful in project learning phase.

Technologies Used

TensorFlow with Convolutional and Recurrent Neural Networks will be useful in project learning phase.
Audacity: https://www.audacityteam.org/
Jupyter Notebook: www.jupyter.org/