This demonstration uses AT&T's Watson Speech recognition engine and AT&T Natural Voices synthetic speech in an Angular.js javaScript MVC framework.
Push and hold the Press To Talk (PTT) button and say phrase that is consistant with the selected grammar. For an MGRS Grid location such as 17CMM12345678, say "one seven romeo mike mike one two three four five six seven eight" and then release the PTT button. You will get better results if you use good PTT technique. Push the PTT down, wait a half second, say the phrase, wait a half second, then release the PTT.
If all goes well, you should see the recognition results in the table and hear the computer repeat the hypothesis back to you. To get good recognition results, it helps to have have a very good audio source. Headphone microphones are the best. Make sure that your audio levels are good. Too low and the recognizer will have difficulty distinguishing between the utterance and the background noise. Too high and the signal will be overloaded or distorted.
The first time you run this demo, you will get a warning as the program is deploying a Java library that is used to send the audio to the server. There may be a second warning the first time you push the PTT. Agree to deploy Java and to not block the application.
Try out the syntetic speech. Type a phrase in and hit play. You can change the voices too. Rosa and Alberto speak Spanish, so type the prhase in Spanish if you want it to sound right. Try "Hola, mundo."