Test utterance detection against synthetic data (as unit tests) and annotated wav files (functional tests).