One way might be to run the audio stream through a speech-to-text engine and parse the resulting transcript. A video recognition system could also be used to i

One way might be to run the audio stream through a speech-to-text engine and parse the resulting transcript. A video recognition system could also be used to identify faces, landmarks and common objects.