Signal-processing of audio for speech-recognition