The Google logo is displayed on a sign outside of the Google headquarters in Mountain View, California.
(Photo by Justin Sullivan/Getty Images)
Google is perfecting its DeepMind system by using artificial intelligence (AI) generated voice.
Advertisement
Google’s DeepMind
According to Android Headlines, the growing use of virtual personal assistants such Amazon’s Alexa, Microsoft’s Cortana, Apple’s Siri and Google assistant, AI speech is becoming more important. Currently, the virtual assistants use speech that is actuallt real human speech rearranged as necessary. This relies on pre-recorded words and phrases and it is called concatenative speech synthesis.
Google Deep Mind is an AI system that builds learning algorithms by using deep learning technologies and neuroscience. The system aims to create advanced AI that can help to create advanced computers and mobile devices. DeepMind technology was developed initially by an English company, but it was later purchased by Google in the year 2014.
Now, DeepMind will use AI in order to generate its own voice. The system still uses real human speech, but only as a mean to learn intonations and patterns. Then, DeepMind is forming its own speech by using human speech linguistic information. This way, the system will be able talk with sophisticated fluency.
WaveNet
According to Pulse Headlines, the program that will allow DeepMind to generate its own voice is called WaveNet. This system will make it possible achieving a more human speech that is completely fluent. From the same information, WaveNet can generate a variety of different voices. Since it isn’t limited by the available recordings, this system called parametric speech synthesis is potentially much more flexible.
The WaveNet system has been tested by Google on both Mandarin Chinese and English listeners. The system was judged by listeners to be much more realistic than the other speech generators testes.
Advertisement
The talking feature was important for DeepMind, since it is a crucial aspect of communication between the human user and the machine. In its aim to make WaveNet talk like a person, Google created new technology beyond the text-to-speech systems (TTS). The WaveNet human-alike voice system is not practical yet in real-life devices, since it is still in a beta phase.
