Microsoft SilentVoice System of Recognition of the Silent Speech

2018: Announcement

In December, 2018 Microsoft provided the system of recognition of the silent speech which, according to in the company, works very precisely.

In a system under the name SilentVoice are used the special microphone with the pop filter and the software. To activate recognition, it is not necessary to say any commands — to articulate words enough, without using vocal chords: the technology defines a difference in the air flows created by different types of the speech.

According to developers, the new interface of voice input allows to recognize the speech even if speaking does not say a sound, with a 98.8 percent accuracy. A system is capable to distinguish sounds with a volume less than 39 dB. It fixes fluctuations of air near the user's lips. A system decrypts them  and  transfers to  the heard voice or the text. The technology cuts extraneous noises and  records a voice very purely. Sound recording is performed at a breath, but not on an exhalation.

Microsoft says that new development is capable to solve a problem of the complicated use of voice assistants, phone conversations and record of voice messages in places of accumulation of people, such as public transport and open office space.

The company also calculates that SilentVoice especially is useful to wearable electronics: the "smart" microphone can be built in the computerized watch, a ring or a Coulomb.

Microsoft provided the system of recognition of the silent speech with a 99 percent accuracy

Possibilities of SilentVoice were shown at a conference on problems of a human factor in   ACM CHI 2018 computing systems.

Microsoft works on voice recognition systems, however, as well as competitors long ago. So, in October, 2018 Google presented VoiceFilter technology which by means of artificial intelligence isolates a voice of the specific person among noise.^[1]

Notes