A deep neural network was built using TensorFlow and Keras API.
Under normal environment, audio signals from 74 children were collected with each child reciting various syllable sounds.
Sounds were segmented using WaveSurfer app and their MFCC features were derived using Librosa which is a python package for audio analysis.
The various syllable sounds that were analysed are listed below:
| Root letter | Sounds |
|---|---|
| B | ba baa bi bii bo boo bu buu bai bau |
| D | da daa di dii do doo du duu dai dau |
| G | ga gaa gi gii go goo gu guu gai gau |
| M | ma maa mi mii mo moo mu muu mai mau |
The dataframes for the features of the syllable sounds were created using Pandas.
Data of 60 children was used for training and 14 children was used for testing.