The winner of the Assistive Tech track at HackNYU 2017. Implemented in Python and Processing using Computer Vision and Natural Language Processing modules.

YouTube Link for the Pitch:


When it comes to Cerebral Palsy and other potential diseases that causes immobility and speech impediments, the options that are available to give someone voice are often expensive and out of reach for many. Because of this, people with this condition have often had to rely on head mounted lasers, a print out, and the assistance of others to speak their voice. We wanted to change this and provide a low cost option that would turn voice to vision- thus introducing VOICE[H]OVER.

What it does

VOICE[H]OVER works much like the head mounted laser and print out, but is notably different because of its ability to export inputs into a .txt file that can be read by a Python code that will speak with natural language. When used effectively, it can be used to eliminate the necessity of help from others. It is unlike other options of this technology because it is affordable and only needs a computer with a webcam.

How we built it

We built VOICE[H]OVER primarily using Processing and an Open CV to track the laser. This program then converts the inputs into a .txt file that can be read by a Python Script attached to natural language processing.

Challenges we ran into

Although we worked tirelessly to include completed options of eye tracking and predictive text, we could not get this into the final state. To show our development, however, we included this code as well, so you can see the progress we made. We even have a fun, though challenging, eye tracking game you can try!

Accomplishments that we're proud of

We are proud that our friendship is still in tact ;) and that we were able to get as far as we could when we didn't think even half of this was possible.

What we learned

Don't think too big. Focus on the main goal and take it step by step. Know what goals are stretches and what is more realistic.

What's next for VoiceHover

To include data stability, eye tracking, predictive text, and an overall better design.