An audio version of the filter bubble
Hearing aids would be more efficient if they were capable of filtering conversations to focus on the speaker who interests the user. This is now possible, thanks to a prototype developed at the University of Columbia. In a multi-speaker environment, the device detects which person the user is trying to listen to, based on their neural signals, then uses a neural network to separate and amplify the right voice.
This innovation is a fluid cooperation system between humans and artificial intelligence, in which the person does not even need to express their intention. This hearing aid, obviously precious for the hearing-impaired, could turn out useful for everybody - to avoid having to prick up your ears to understand a conversation in a family dinner, or to eavesdrop on the diners at the next table!
Spotted in 2018
- United States