imaduddinamajid / baby_cry_detection

Recognition of baby cry audio signal

Baby cry detection

Recognition of baby cry from audio signal

The aim is to automatically recognize a baby crying while sleeping. In such case, a lullaby is started to calm the baby down.

This is done by implementing a machine learning algorithm on a Raspberry Pi.

The code is organised according to following schema:

• TRAINING

  It includes all the steps required to train a machine learning model. First, it reads the data, it performs feature engineering and in trains the model. The model is saved to be used in the prediction step. The training is performed on a powerful machine, such as a personal computer.

• PREDICTION

  It includes all the steps needed to make a prediction on a new signal. It reads a new signal (9 second long), it cuts it into 5 signals (5 second long), it applies the pipeline save from the training step to make a prediction. The prediction step is performed on a Raspberry Pi 2B.

Part of the data used for training comes from ESC-50: Dataset for environmental sound classification