A team from the Mitsubishi Electric Research Laboratory in Cambridge, Massachusetts have unveiled a new AI, which they say can overcome what has been dubbed as the "cocktail party effect" and select a single voice from a group.
There is thought this could be applied for use with voice recognition services like Amazon's Echo and Google Home to provide customers with an improved service.
The software was trained using 100 different English speakers, but will be able to detect different languages too.
Niels Meinke, a spokesperson for Mitsubishi Electric, said: "It was trained using 100 English speakers, but it can separate voices even if a speaker is Japanese."
Meinke has revealed the system is able to detect and separate two people speaking into one microphone with an accuracy of 90 percent. Whilst for three speakers, the accuracy is still at an impressive 80 percent.
The team at Mitsubishi Electric Research Laboratory also believe this new AI could improve automatic transcriptions as well as help fight crime by being able to separate music, for example, from a person speaking more easily.
Meinke told New Scientist: "The system could be used to separate speech in a range of products including lifts, air-conditioning units and household products."