Speech recognition came a long way since its tentative beginning a couple of decades ago. At first, Dragon Naturally Speaking, created by Nuance, was a big and expensive software package that users had to load onto their computers. Once installed, a user could dictate entire documents into Microsoft Word. Of course, it was slow and inaccurate, and it had trouble with different accents, making the whole thing a frustrating experience. Worse still, anyone working in an office alongside other people could not use it because everyone could hear the dictation. It eventually became part of the Microsoft package but is not likely to work well in today's open office plan.
When smartphones took over, our society became a little more accustomed to disembodied conversations being audible anywhere and everywhere. Even the sight and sound of people apparently talking to themselves (they have Bluetooth headsets on) became commonplace.
Then voice assistants came along and added to the cacophony. With Google Assistant, Siri and other such tools becoming more entrenched in everyday life, there would be little point if one could not be free to voice-command anytime, especially when there is a specific need to remain hands-free. But try telling your favourite assistant to add diapers to the shopping list or give a colleague's work a quick look-over, and you are bound to get a few odd looks. Microsoft's suggestion: Whisper your command.
The Redmond giant has patented a solution called SilentVoice and describes it as a "new voice input interface device that penetrates the speech-based natural user interface in daily life." The proposed ingressive speech method - it requires you to inhale while whispering instead of exhaling as we usually do - places a microphone about two millimetres away from the mouth so that it can capture the ultra-small voice leakage and process it. The technology ensures a good signal-to-noise ratio, which means background noise would not distort the whispered voice command.
You get a double benefit: First, no one can hear those barely audible whispers and there will be no privacy breach; second, unlike people shouting in their phones, you will not annoy anyone whether you are at home, in the office or commuting. Finally, the service is quite reliable. By measuring airflow direction, SilentVoice can separate your commands from usual utterances with 98.8 per cent accuracy and no activation words are needed.
Microsoft has a video demonstration in which a researcher shows the use of a little handheld device comparing normal speech with whispered speech. Although humans cannot catch the whispered words, those can be picked up by the microphone and even transcribed into text. The day may not be far when people will be walking around with gadgets firmly pressed against their lips as they breathe in their words.