Speech recognition is one of the foundations of artificial intelligence that is slowly coming into our lives, especially thanks to today’s smartphones. In fact many of the advances we are seeing lately have to do with the management of devices via voice. And for this, it is vital that we understand the phone perfectly. For this reason, Microsoft has improved the voice recognition system on their handsets with Windows Phone 8 operating system.
For now this Microsoft speech recognition can be enjoyed by only users of Windows Phone 8 living in the United States and to use Bing, as this feature is only implemented in the form of Microsoft. And say for now, because we assume that also hit Europe soon. What seems more likely is that this improved voice recognition reaches the terminals with Windows Phone 7.8.
The improvements primarily consist of 15 percent reduce recognition errors, doubled the speed at time to recognize what has been said and also much improved in noisy environments, perhaps the Achilles heel of most voice recognition programs.
Deep Neural Networks (DNN)
Redmond’s company has used deep neural networks (DNN) in order to develop a more efficient version of its speech recognition model. In this way the computer system is copying our brain works when it comes to recognizing what we say. Like us, this computer model also requires training by scientists working at Microsoft, as the DNN contains hundreds of millions of parameters representing the human voice patterns.
The success of the new Bing voice recognition in the terminals with Windows Phone 8 reaches 85 percent, which, though still far from 100%, to which humans sometimes not arrived, if that can be considered a significant approach. In addition, the increased speed will cause every time we talk more naturally with our phones, since your response is virtually instantaneous.