Learning to speak
gunthercox opened this issue · comments
A while back I had an idea to set up an experiment to try to improve the quality of speech provided through a simple Arduino text to speech library: https://github.com/jscrane/TTS.
Abstract
Google has awesome speech recognition capabilities, an example can be seen at https://www.google.com/intl/en/chrome/demos/speech.html. My idea is to set up a system that iterates through a list of words such as a .txt file version of an English dictionary. Each word would be sent to the Arduino board to be converted to speech. Google's speech api would need to listen to each word and return the text that it thought the spoken word was. A program would need to be written which checks to see if the word sent is the same as what was detected. If the words match then it should continue to the next word, otherwise it should make some small modification to the TTS library's sound bytes, recompile, and then download them onto the Arduino board. This is basically a brute-force method for making the text to speech library clearer, however it would be very interesting to see if this would actually improve the quality of sound produced.
Speech sample
Sometimes even the wisest of men and machines can be in error.
Current speech sample: [link]