For running the code locally you'll need python and jupyter lab. However we recommend using colab as it already has the setup.

• data_processing.ipynb: This file is used for data visualization and extracting various spectrogram features from the speech.
• baseline.ipynb: This file contains the code for training baseline models such as XGboost adnd Naive Bayes over the spectrogram features. For XGboost feature importance can alos be visualized.
• wav2vec.ipynb: This file contains code for extracting wav2vec features from the speech. Further a Multi Layer Perceptron is trained on top of the extracted Wave2Vec features.
• PhoneticClassifier.ipynb: This file contains code for training a CNN architecture on the spectogram of speech. As a proof of concept the model is trained on speech signals containing sh phonemes. Multiple phonetic graders can be trained on different phonemes. The output of these phonetic graders can then be ensembled to get fina verdict.

XGboost - https://xgboost.readthedocs.io/en/stable/
Wav2vec - https://huggingface.co/docs/transformers/model_doc/wav2vec2
Scikit-learn: Machine Learning in Python, Pedregosa et al., JMLR 12, pp. 2825-2830, 2011.
Librosa - https://librosa.org/
pickle - for saving data https://docs.python.org/3/library/pickle.html