Inspiration
We wanted to make interesting application using voice analysis that can distinguish human's voice!
What it does
This program randomly selects a celebrity then user will try to mimic the voice with mic. The program will return the accuracy score that indicates how similar it is to the actual celebrity voice.
How we built it
We used Flask to link JS to trigger Python program. Also the analyzer is modeled by machine learning approaches.
Challenges we ran into
The machine learning model does not fits well. Therefore, we had to do fine tuning + normalizing the data to train the model. Also audio clip collections were tough. Since we did not have much knowledge on machine learning we couldn't fully finish the project.
Accomplishments that we're proud of
We are most proud of is that everyone engaged and get to try 4 different ml models.(RF, MLP, KNN,SVM).
What we learned
We get to learned lots of different approaches in training machine learning models. Also learned the risk of high overfitting and importance of number of dataset.
What's next for Voice Score
Need to improve accuracy in our training models.
Log in or sign up for Devpost to join the conversation.