Listening to a single voice from a group of people who speak simultaneously is a challenge that most of us can overcome with minimal effort, and it is enough just to look at that person to discern the words spoken by the ambient noise. The same can not be said about equipment that uses voice command, their algorithms not being able to differentiate user instructions from the words spoken by others in the room
Trying to find a solution to this surprisingly complicated issue for a computerized system, Google has recourse to artificial intelligence technologies to mimic what people do virtually effortlessly, namely to identify and isolate voices from the crowd on only that person in time what he is talking about.
To demonstrate the effectiveness of the new AI filter, the developers team used the scenario of a comedy show in which two participants speak simultaneously while the audience acclaimed in the background. Reduced to a simple left-right adjustment, the filter can divide the sound into distinct sound strings, one for each voice identified in the image. It is remarkable how the ambient noise is entirely canceled, and the selected voice is preserved even when the listener partially covers his face by gesturing his hands
Surely, the applications of this technology are multiple, augmenting surveillance cameras with advanced listening capabilities of filmmakers is just one of the possibilities. But most likely, Google plans are rather harmless, only to improve existing messaging services such as Google Hangouts and Duo
Implementing a Voice Splitting Software algorithm can also improve the use of voice support services, better distinguishing the words spoken by ambient noise. However, technology could put in place organizations that guard over the right to privacy and private conversation in public spaces, abusive use becomes hard to prevent if any smartphone or camcorder will have the voice separation feature implemented as a standard