When a train rattles by, it is often very hard to understand announcements in the train station. Researchers at the Fraunhofer Institute for Digital Media Technology (IDMT) in Oldenburg, Germany, have developed a software that significantly improves speech intelligibility – it works for loudspeaker announcements, for the voices of speakers at conferences or conversations on mobile phones.
The software uses a microphone to analyze the background noise and adjust the speech in real-time. It is not sufficient to just make the voice louder. Instead, algorithms prioritize certain frequencies and selectively boost those that are particularly disturbed by the ambient noise. As well, loud parts of speech are systematically subdued while quiet parts are amplified. So you always know where to catch your next train.