Google Duo is using machine learning technology to better the quality of calls.
The internet calling platform is using the Machine Learning Technology (MLT) from DeepMind, its neural network technology.
Google focuses on the new feature to address customer concerns regarding audio jitters- the universal internet calling menace.
Data from the call sender is usually broken down into packets before travelling to the receiver. The packets must be assembled in the best technology possible for the receiver to get a high quality stream of audio or video.
Temporary network glitches sometimes lead to loss of packets under transmission. Although packet loss concealment PLC is used to fill in the missing packets, the technology is not 100 percent effective.
It is due to the glitches that the new Google WaveNetQ PLC system has been set up to improve call quality.
WaveNetQ will use machine learning to replace most of the lost packets with artificially generated packets with human-like sounds.
Making of calls in places with noisy backgrounds also finds a new and better solution, WaveNetQ will augment data and cut the background noises.
Currently, about 99 per cent of all calls made on due face packet loses due to network glitches and excessive jitter. This is almost all the calls made on Duo.
The company has reported that it is using over 100 speakers in over 40 languages to train WaveNetQ on production of human-like sounds.