I think as silent budha, Ms states that Natal could works with some "pre natal" games, the games can be patched to handle the new input but it's unlikely have the cpu ressource do deal with the amount of data generated by Natal.
a photo of Roku Player:
If Ms want to make the most of their take they have to free as much as they can the 360 cpu.
Natal may end with quiet some work on its hand, all together:
generate 3d squeleton and track its motion
2D shape recognition
sound processing
I would be surprised if the system pack an ARM/MIPS tiny/cheap tied to some kind of DSP + tiny amount of ram and rom.
Mobile phones have word recognition for years. You need to do some funky maths to get numerical data for waveforms that lets you do comparisons, but it's DSP type stuff which Cell is very well suited for.Well as to CPU power who's to say. But conventional speech recognition still requires a fair bit of CPU power to recognize and track just one person speaking.
A quick Google.The Natal goes one bit further in that it can recognize and keep track of multiple speakers presumably speaking at the same time. And not only that adds in background noise removal and presumably echo cancellation so that you can use it instead of a headset for voice chat while gaming. I wish the PC had something similar.
The video stream thing is a webcam's abilities. It just sends a honking great bitmap to the console. (note:more on this later...)Then additionally it has to record and process 2 distinctly different video streams and then integrate them into something that can represent what it see's in 3 dimensions.
When you've mapped the skeleton, following out-of-sight limbs comes with the system with no extra overhead.And not only that the claim is that it can recognize and differentiate between multiple people. As well tracking items/people/apendages that may pass out of sight of the camera due to being occluded due the movement of other people or the objects.
Again, mobile phones and cheap cameras have this tech.And on top of all of the above it also has to be able to recognize facial features (enough to distinguish one person from another) as well as track changes to those facials features (smiling/frowning). And likewise do it for each and every person in the field of view.
Well, there'll definitely be some processing overhead on PS3, but it may well have enough capacity to swallow it up without any impact on the game. A lot depends on what exactly Natal is doing. Having seen the report on the box size, there's more in there than I guessed, and there may well be more processing going on then I expected, and onboard RAM etc. We don't know what data is being passed to the console. Is it just a couple of framebuffers? Is it accompanied with skeleton data? Is it multiple audio streams or just one feed of Fourier tranformed audio data? At this point we're guessing!I dunno, maybe all of that takes virtually no memory or CPU power. But it at least appears as if it would take some non-trivial amount of processing power and memory. Memory BTW - that both consoles are fairly short on. Considering to do something similar with the Eye Toy + Wands the PS3 would have to do all the work as well as use system memory.
If I got it correctly, the NATAL device itself should include a powerful enough processing unit and memory separate from the console in order for the 360 to run satisfactory demanding games with all the features promised. Right?Well as to CPU power who's to say. But conventional speech recognition still requires a fair bit of CPU power to recognize and track just one person speaking.
The Natal goes one bit further in that it can recognize and keep track of multiple speakers presumably speaking at the same time. And not only that adds in background noise removal and presumably echo cancellation so that you can use it instead of a headset for voice chat while gaming. I wish the PC had something similar. Using Teamspeak without a headset is begging for some horrendous background bleed that's enough to make your ears bleed as well as everyone else in the Teamspeak channel.
Then additionally it has to record and process 2 distinctly different video streams and then integrate them into something that can represent what it see's in 3 dimensions. And not only that the claim is that it can recognize and differentiate between multiple people. As well tracking items/people/apendages that may pass out of sight of the camera due to being occluded due the movement of other people or the objects.
And on top of all of the above it also has to be able to recognize facial features (enough to distinguish one person from another) as well as track changes to those facials features (smiling/frowning). And likewise do it for each and every person in the field of view.
I dunno, maybe all of that takes virtually no memory or CPU power. But it at least appears as if it would take some non-trivial amount of processing power and memory. Memory BTW - that both consoles are fairly short on.
Considering to do something similar with the Eye Toy + Wands the PS3 would have to do all the work as well as use system memory.
Perhaps that's why it hasn't been demo'd with a game that could possibly stress the system. IE - retrofitting it for use in a conventional game. After all neither a painting application nor the Eye Pet appear to be graphically challenging or memory intensive.
Regards,
SB
it appears Sony's motion-sensing efforts may be farther along than many thought. Speaking with GameSpot, Sony Computer Entertainment America senior vice president of marketing Peter Dille divulged that the technology--first revealed in a US Patent filing last fall--is already in developers' hands.
"We're a little bit past the research phase," he told GameSpot. "We're having conversations with the third-party community. The dev kits have started to go out to the third parties as well. They're working on the tech. They couldn't be more excited about it."
and look at 59mn. Richard Marks have also made some demo with Depth camera like project natal few years ago (2004)
http://stanford-online.stanford.edu/courses/ee380/040121-ee380-100.asx
maybe Sony have a project like natal in their box for PS4
Richard Marks at Siggraph 2000
http://www.research.scea.com/research/pdfs/SIGGRAPHetech2000.pdf
not true 1:1 and good accurate but near concept than controller shown at sony conference
tech borrowed from some Israel compary
My presumption, based on joker454's comments, is that EA Sports is already working on PSMC stuff. I think it will be very easy for them to have PSMC integrated into their PS3 versions of Tiger Woods and Grand Slam Tennis by next summer.
I don't see how that holds, Microsoft's solution has the least compatibility with the standard (Wii-mote). Natal games will have to be very specifically designed for it, control schemas for Wii-mote/Waggle targeted games will simply not fit it ... unless they compromise and include a wand.
Are there any Sony studios we haven't heard from lately?
If anything, I see the opposite.
There is one crucial point that is often missed though, and I alluded to it earlier. Natal is a complementary technology. It can be used in conjunction with all their existing control systems, and from everything microsoft has said, it will not significantly impact performance on the console. (From what I can tell, the demo they were showing behind closed doors was *entirely* processed externally).
It's this combination which I think will be the critical point. A set of well developed systems (motion tracking, facial/image recognition, voice tracking and recognition). With potentially little to no overhead and the ability to complement an existing control scheme. For developers, it becomes a question of what they can add to improve instead of what they can replace to improve, which I feel is important. If this process is as easy as they have implied, then the success of the system is far more likely (in my view).
If anything, I'd compare it to the innovations the DS introduced. Touch screen, microphone and a second display. It didn't limit any existing style of game (after all, all the controls were still there) however it did open the door for all sorts of unique and unexpected games - but importantly practically every style of game could take advantage of at least one to improve the game some how.
It will be interesting to see what odd ball uses developers come up with. I would have never even considered the milo/tamagotchi angle.
Two things that come to my mind, are the old Wii hack point-of-view demos are now possible (also TrackIR style interaction), also in theory it would be possible to mix the video stream and face tracking to put your face in a multiplayer game - in real time. There are loads more, but these two things are good examples of things that don't put huge demands or requirements on the game design.