HappySqurriel said:
Reasonable said:
HappySqurriel said:
Reasonable said:
To be honest I'm more than worthy. If anything I'm wondering why these things take so long to arrive. The tech for such stuff has been around for quite a while with decent SW support too. As ever its only when someone makes money with innovation (hello Nintendo) that you see it get the full push from all parties. |
Depending on who you talk to, the technology to produce NATAL well hasn't been around for long and may not even exist yet ...
The digital sensors that are used to capture videos are (from my limited understanding) not different from the ones that are used to capture still images and when they capture videos you only use a small portion of a much larger sensor to capture a particular frame, and over time (in about a second) you use the entire sensor and cycle back to the first frame. This approach has been used to capture images at 120fps at very low resolutions and is currently being used on state of the art digital camcorders to produce video at 720p/1080p at 30fps.
Now, I don't know the technical details, but I suspect that NATAL captures images at below HD Resolutions (probably very close to SD) at 30fps in order to keep manufacturing costs down. It probably doesn't have the best lenses, and a person at 6 to 10 feet away from the camera is probably pretty blurry and has some distortion (like you would see from a web-camera). On top of this, processing power alone limits their ability to track more than a couple dozen datapoints, which means that the system can focus on the big picture for a person (there skeleton) or small details (hands or face) but probably not both at the same time; and certainly not both at the same time for multiple people.
5 years ago the technical limitations of an affordable product prevented it from being produced, and 5 years from now the technical limitations will be so minimal that it will be easy to realise a device like this. The question that remains is whether the technology exists today ...
|
All the core elements have been around, in seperate ways, for a while. What we're seeing is the ability, more recently, to couple fairly inexpensive cameras with software that does a decent job of the image/voice recognition with a small resource footprint. I suspect the real challenge facing MS is getting it working very well, and consistently, without requiring much addition resource and with minimal lag due to processing of input through to the actual game (whatever it may be).
As I mentioned in a similar post I've worked on Expert Systems in the eighties from the SW point of view, as well as the use of (admitedly much more expensive) cameras in crowd tracking and flow simulation systems. I've also worked with voice recognition and image recognition. Like I say, I'm not saying Natal as shown has been around (or possible) for ages, but that the core ability and elements have.
|
I was thinking of it from a comercial product perspective more than anything else ... I'm certain that state of the art equipment may have been able to achieve similar results almost a decade ago, but outside of a university lab or a few very rare business applications it (probably) hasn't become cost effective until (very) recently.
|
Yeah, I thought that's where you were coming from. I was talking about capability, not cost or commercial viability. Of course those matter to those wanting to make money from the tech!
