Inspiration

Our collective fear of public speaking. We all have our outlets of nervousness while speaking in front of a large crowd, whether we mumble, speak too quickly, fidget, swing around, sweat profusely, etc.

What it does

This application monitors your speech from multiple approaches and outputs feedback and analytics graphs. It records your speech, but not just with audio -- it also tracks your brain waves and your arm gestures and movements to determine your concentration, boldness, and nervousness.

How I built it

We split up the project into three fronts: the front end, the back end, and the hardware integration end. On the front end, we built a web page with which the user can interface using HTML, CSS, and JavaScript, making use of JQuery and Twittter Bootstrap. The front end communicates the user's requests to the back end, a Java server, which makes use of IBM Watson API's, and Google Charts API's. The server relies on hardware integration code to supply data about the user's movements and brain activity. The hardware integration utilized custom algorithms for interpreting the input data.

Challenges I ran into

For the Myo armband, a huge challenge was to be able to determine between small, rapid movements like unwanted fidgeting or trembling, and decelerating large, bold gestures. We also had generic connection issues with the Muse. At times, we also struggled with understanding the various (sometimes scarcely documented) APIs we had to use.

Accomplishments that I'm proud of

To tackle the Myo armband gesture distinguishing challenge, we used a combination of calculus and thresholding for positioning, and acceleration to filter out these different movements, then smoothed out the data with approximation algorithms.

What I learned

That 36 hours is not enough time to get done everything you want to. But we learned a lot about collaborating, being resourceful, and working under heavy stress.

What's next for How2Speech

Real-time support, and virtual reality features for live feedback from a virtual audience.

Share this project:
×

Updates