Classify infant vocalizations as babble, cry, fuss, laugh, hic.

Author: Yijia Xu [email protected] https://github.com/orgs/uiuc-sst/people/yijiaxu3, 2017

The best classifier was a pretty simple architecture: linear discriminant analysis, applied to a vector of OpenSmile features.

Example of data cleaning and feature extraction make_balance_dataset.py make_tfrecords_FBANK.py

Feature selection feature_selection.py

CNN model CNNmodel.py

Train and evaluate CNN train_prosody.py eval.py

Train and evaluate LDA LDA_test_reliability.py

Evaluate CNN+HMM HMM_CODE/caveneuwirth.py

Related: transcribe mom and child speech, https://github.com/yijiaxu3/speech-segmentation-transcription