FRONT-END SYSTEM STRUCTURE

• Each front-end algorithm converts speech audio data into observation vectors in the probability space.

• A speech recognizer statistically analyzes these acoustic observation vectors to hypothesize message content.