Intel Speech recognition technology does not rely on cloud processing speed will be super Siri

According to foreign media reports, Intel's speech recognition technology does not need to be based on cloud processing and will be faster in response than Siri, according to executives.

The problem with the current speech recognition system is that it's too slow. Anyone who has ever used Apple Siri or Google Voice search must have this experience, and sometimes even simple instructions require a user to wait a long time.

The problem is not that the speech recognition software itself, the accuracy of such software can be said to have reached an unprecedented level. The problem is that speech recognition is still a challenging challenge from a computational standpoint, and the big mobile platforms, whether they come from Google, Apple or Microsoft, must first deliver your voice-recording compression packets to servers hundreds of or even thousands of of miles away. There, computers that are more powerful than mobile phones or game consoles will turn voice recordings into text or instructions. It is this round-trip process that makes speech recognition on most devices very slow, especially for devices that use slower cellular networks.

Since the launch of the Android 4.1 system, Google has been providing offline voice recognition function, but it is still in the experimental phase, not to Google application developers. In addition, Scott Huffman, head of Google's dialogue search department, revealed that while Android could perform some offline processing of voice commands, the operation was far less accurate than the usual cloud approach.

Intel wants to process your voice directly on the phone, not through the clouds.

Mike Bell, the Intel wearable technology director, recently Maik Bell an exclusive interview with Quartz, revealing that the company has a solution. Intel, in collaboration with a Third-party company, integrates the latter's speech-recognition software into Intel's powerful mobile processor, which resolves the human voice and is small enough to be handled without the cloud. The product of this program is a wireless headset for "Jarvis". Jarvis can also act as a voice controller and personal assistant, both to listen to instructions and to respond.

Intel's speech recognition solutions are faster in response than cloud-based competitive offerings and support for offline use.

Speech recognition has been on the desktop for several years, but the phones, other mobile devices, and wearable products are not powerful enough to handle voice commands directly. Intel claims that its technology to build powerful microprocessors for servers and PCs gives it a unique advantage in developing mobile processors.

Mr. Bell said Intel was selling its speech recognition technology to some handset makers to help them differentiate and sell their products.

The result may be that the consumer will be able to use a speech recognition device that can be used for actual conversations. That may mean that it will ask you which "Mike" you mean when you give your phone a simple instruction like "send a message to Mike". People will be expected to use the artificial intelligence dialogue system depicted in Iron Man-turning computing equipment into real personal assistants.

Translator: Lebang

(Responsible editor: The good of the Legacy)