Hacker News Reader

Comment by nelmaven | Hacker News Reader

Someone should tell Google to add speech balloons in their glasses to the people that you're talking too, like in videogames.

That would be revolutionary.

View on HN

Like the author, I'm deaf as well. Real time captioning functionality using an improved version of Google's speech-to-text technology (the current implementation on Youtube is VERY hit or miss) was pretty much my first thought when I saw those glasses.

That's something I've been dreaming about for 20 years or so.

Although this would be awesome, I don't know how we could isolate sound source and make it perceptible to the audio transcriber, on a street with lots of people walking around and talking this would be complicated.

Nevertheless this would be a great idea.

Use a directional mike aimed at roughly one's center of vision?

I feel like that cuts out some of the uses... perhaps you could us ML techniques to separate out different voices and then individually transcribe them?

edit: wow this is a complicated problem now that I am thinking about it... the amount of processing the brain does...

I may be wrong, but I believe Kinect uses the information from its vision system that identifies people, to help inform the microphone array system with data that it can use to identify which of the people in front of it is talking at any one time...

I'm sure we can hack it in there once they're released!

Reply on news.ycombinator.com