Inspiration
A need for asynchronous check-ins by patients, especially in therapy, where a psychological evaluation of the patient must be delayed due to a doctor's busy schedule or absence.
What it does
It takes in a voice recording (in .wav format) and can classify the primary emotion it displays with around 60% accuracy.
How we built it
Used React to make a simple web app with a voice recording prompt, and Python to create and train the neural network.
Challenges we ran into
Building a website to accept and process user input.
Accomplishments that we're proud of
Training a model from over 12,000 audio files, generating over 2 million unique points.
What we learned
LSTM (Long Short-Term Memory) and neural networks concepts, as well as some Django and React.
What's next for AELM
Building a usable web interface; optimizing the model.
Log in or sign up for Devpost to join the conversation.