This repo contains scripts on how we can interpret baby cries based on donateacry-corpus. (https://github.com/gveres/donateacry-corpus)
- Infant emotion PREDICTION : Our corpus contains five folders each of them represents an infant emotion class.
- Belly_pain : 16 wav files.
- Burping : 8 wav files.
- Discomfort : 27 wav files.
- Hungry : 382 wav files.
- Tired : 24 wav files.
All of these sound files lasts between 6 and 7 seconds. This data is not balanced and the Hungry class represents over 80% of the whole data. Our goal is to maintain a balance between these classes by collecting data or by other data augmentation techniques for a better modeling.
We have used a deep learning approach to predict the sentiment.