A flask application to predict the speech emotion of any .wav file.

This is a speech emotion recognition app. It will allow you to train a modular MLP model with the RAVDESS dataset, and then use that model with a flask application to predict the speech emotion of any .wav file.

REQS:

To download the RAVDESS speech emotion recognition data, go to: https://drive.google.com/file/d/1wWsrN2Ep7x6lWqOXfr4rpKGYrJhWc8z7/view

for installing all dependencie simply open terminal and run:

. ./install_deps.sh

This should create your venv and populate it with all necessary dependencies

MODEL:

A multilayer perceptron model to detect the emotion of wav files. To create and edit the model see create_model.py Once the create_model.py is adjusted to your liking (emotions_to_observe, and path to sound data), simply run:

python3 create_model.py

to create the model.model binary file and test accuracy of your model

APP:

Once the model.model binary is created, you can spin up the flask application (ToneCheck): To do so run

. ./start_flask.sh

The app will run default on localhost:5000, the emotions available for predictions will correspond with the emotions_to_observe variable you have edited inside create_models.py (and are therefore available inside the model binary file)