Arsh Dilbagi wants to help the one hundred million people around the world with limited speech capability. His project, winner of the Voters Choice Award in the 2014 Google Science Fair, is Talk. Talk uses a small sensor and a MEMS microphone to convert breath to speech.
Talk's non-invasive sensor can be worn like a headset and sits just under the nose. A secondary configuration can also be used to catch breath from the mouth. Using short and long breaths the controller takes the input and uses an algorithm based on morse code to convert to words and sentences.
https://www.youtube.com/watch?v=3H73oYijLeg#t=11
Nine different voices are available for the user to choose - male and female voices are represented along with different accents. The processor was designed to be the same size as a smart phone and fit in the user's pocket.
Project goals were to create a faster speaking rate and make the device more affordable than current augmentive and alternative communication devices (AACs). Rough calculations say that the speaking rate using Talk can be increased by up to three hundred percent over commercially available devices.
Like many Google Science Fair winners Arsh has an incredibly detailed discussion of his method on the project page. Prototype iterations are shown for the makers and programmers can review the different algorithms and inputs for the breath to speech conversions.
Raw data is also displayed on the page, showing the range of long exhales vs short exhales for nine different test subjects. Testing was done by Dilbagi, his family members and eventually a Parkinson's patient at Sir Ganga Ram Hospital in New Delhi.
Talk is another excellent example of a student who has a clear goal in mind and uses available technology to find a solution. Having design constraints of speech conversion time and cost as his metrics and rigidly following the scientific method also solidify this as a fantastic device.