Microsoft Speech Recognition Hits Speed Track: Going Faster and Accurate for Windows Phone

June 18, 2013, By Sanjeev Ramachandran

Who doesn’t know Microsoft has been trying hard to beat rivals Google and Apple in the OS scenario. A fair amount of success has also come its way in recent times. As if not to rest on its laurels, the company’s research unit is now going for more!

The latest on the block is a new technology that promises to help improve voice recognition and speech features in Windows Phone devices.

Bringing to the fore this new addition is Microsoft’s Bing team, who has announced that voice search and voice-to-text features in Windows Phone are now twice as fast as in the past. The engineers in the team have claimed that the accuracy of these features have been improved by 15 percent.

Windows_Phone_8_event_03

According to Microsoft’s Bing team researchers, this new technology comes thanks to a computation model that mimics the way human brain works. Now that is awesome, indeed.

If you are looking deeper into it, see this is what we found. Folks at CNET say that some kind of deep neural network technology has been applied to speech recognition so that it further helps users to search or compose text message with their voices.

To implement speech recognition features in a computer device may not be that easy. It is difficult for a computer device to recognize voice in real time, especially in noisy environments.

Moreover, when the device needs to translate speech into text, the whole process is almost spontaneous. For a smartphone, it might take just one or two seconds to convert voice into text. Errors could be natural and would depend on various factors, such as the background noise and the quality of the microphone in your smartphone device.

Putting an end to all such doubts, Microsoft has claimed that by using this new speech recognition technology, the word error rate could be reduced to 13.5 percent from 16 percent. Contemplating on that, you may want to run through this document done some time ago.

What does that mean to a normal mobile phone user? The accuracy and speed gains may seem very much disappointing. But for those who work in the speech research, this is a significant development and it promises other speech related applications such as real time language translation too.

All said, the new technology could, in future, let you translate your voice into some other language. By the way, we saw that sometime ago, when it was reported that the English word could be translated into Chinese Mandarin.  The video I have for you here from Microsoft Research would throw ample light on that.


However, we needn’t look forward to that happening in a large scale at least in the near future. Till then we will be able to enjoy a much improved voice recognition feature in our smartphones, because Microsoft has already replaced the acoustic model in their speech recognition technology and even started rolling out the update to data centers for the customers in the United States in April.

Though the tech giant has not revealed the exact time of roll out of the update internationally, it is expected to complete the process within the upcoming weeks.

Awesomeness seems to be on its way. What do you have to say about this?

© 2008-2012 DeviceMag.com - All rights reserved | Privacy Policy