Microsoft has claimed that its development division has devised a method of improving voice recognition in smartphones, Digital Spy said.

The computing giant says that the new system is capable of processing vocal commands at twice the speed of current-generation devices by mimicking human brain function.

The method, which is based on deep neural networks, was used to compose text messages and search the internet via Bing, with the results allegedly showing a 15% increase in accuracy.

"For a normal sentence, you will have one less word to correct," said Michael Tjalve, a senior program manager with Microsoft's research division.

Microsoft has refused to compare the system to other voice-recognition software, such as Apple's Siri and Samsung's S Voice, pointing out that other companies have taken different approaches to processing vocal information.

The technology is already being distributed to data centres in the U.S., but plans for an international rollout are yet to be confirmed.