I am trying to make a Java application for live speech alphabet recognition by using Sphinx-4. I will only need the program to recognize a small number of words (i.e, the English alphabet).
I have managed to get the program running using the LiveSpeechRecognition and en-us dictionary. However, it is very slow and inaccurate (~50 x audio time). I am looking into making my own customized .dict file in hope of it running faster and more accurate.
May i ask whether that will work? If so, how can i make my own customized dictionary? If not, how can i improve the speed of the program?