Implement speech to text in AR-glasses

I'm working on a project where I try to use a speech-to-text software together with the AR-glasses Moverio BT-300 and Vuzix Blade (Both Anroid 5.1.1), so that the glasses will be able to show live captions of e.g. a conversation.

I'm wondering about the best architecture choice for this kind of system. Should I use Unity to record speech through the microphone and then send audio and receive text to/from the websocket used? Or should I use an external program for this? If I use an external program, how will this work when using the application on the AR-glasses? I'm very much a newbie at this so any help would be much appreciated!