VoicePd

Application
training

Inspiration

Ive had many people in my life like my grandpa who has had parkinsons and they would have benefitted greatly if they had access to early detection.

What it does

My project uses machine learning to analyse a spectrogram of a voice recording to determine whether the individual who recorded the voice recording has Parkinson’s or not via vocal biomarkers

How we built it

We used the mpower dataset provided by synapse to train a machine learning model with a Efficientnet + LSTM architecture.

Challenges we ran into

Data was a huge challenge. We spent weeks finding high quality and accessible data and waited more weeks to get access to them. Another major challenge was computational power as the mpower dataset was extremely large but the cloud computing service I was using didnt provide enough computational power to process all the data.

Accomplishments that we're proud of

The model cam out great and the fact that the project even came to life after so many challenges was a huge accomplishment

What we learned

We learned how to use audio data to train a machine learning model and learned alot about the more niche and specialised parts of machine learning

What's next for VoicePd

If this project is successful I wish to reach out to universities or companies to provide me with more computational power to create a more accurate model and work on similar impactful projects in teh future without limitations

THE DEMO IS IN THE DESCRIPTION OF THE VIDEO LISTED