4 ms·
Voice processing through the cloud is one way to do it. It allows to perform the CPU consuming part of Voice Recognition on remote powerful servers but to do so
by koopajah 15y ago
Voice processing through the cloud is one way to do it. It allows to perform the CPU consuming part of Voice Recognition on remote powerful servers but to do so you must be connected to internet, it's harder to include your "local" info (own contacts, own music, etc)
With a chip powerful enough you can do the voice recognition directly with your phone as it is done in a lot of devices embedded currently in cars for contact/music recognition.
- ConstantineXVI 15y agoNot necessarily, all the remote servers have to do is churn voice into text. Your phone can handle the sorting through your local data just fine once the voice is processed.
- koopajah 15y agoOf course, you can do Speech to text and THEN analyzes the text to find the "intent", but on embedded devices it is easier to "plan" what is the vocabulary that you want to support and then perform recognition based on this vocabulary. At least on low CPU/RAM embedded devices.
- adambyrtek 15y agoOnly when the space is limited and enumerable, like on legacy GPS devices that keep a complete routing database. Android voice actions are much more powerful, allowing you to search for any phrase, get directions to a given place, send arbitrary text messages etc. The possibilities are effectively infinite.
- biot 15y agoExcept that your data will get split between local and remote. "Book a return flight to visit mom the next time a flight is under $250". Either your device needs to know all about flight times or the remote server needs to know where mom lives, even if she moves. Throw everything into iCloud and there's no need for local processing -- either the audio of your speech or the transcription of your speech is sent to the remote server and everything is done there.